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> | |
f46b5a66 CH |
36 | #include <linux/compat.h> |
37 | #include <linux/bit_spinlock.h> | |
cb8e7090 | 38 | #include <linux/security.h> |
f46b5a66 | 39 | #include <linux/xattr.h> |
7ea394f1 | 40 | #include <linux/vmalloc.h> |
5a0e3ad6 | 41 | #include <linux/slab.h> |
f7039b1d | 42 | #include <linux/blkdev.h> |
8ea05e3a | 43 | #include <linux/uuid.h> |
55e301fd | 44 | #include <linux/btrfs.h> |
416161db | 45 | #include <linux/uaccess.h> |
f46b5a66 CH |
46 | #include "ctree.h" |
47 | #include "disk-io.h" | |
48 | #include "transaction.h" | |
49 | #include "btrfs_inode.h" | |
f46b5a66 CH |
50 | #include "print-tree.h" |
51 | #include "volumes.h" | |
925baedd | 52 | #include "locking.h" |
581bb050 | 53 | #include "inode-map.h" |
d7728c96 | 54 | #include "backref.h" |
606686ee | 55 | #include "rcu-string.h" |
31db9f7c | 56 | #include "send.h" |
3f6bcfbd | 57 | #include "dev-replace.h" |
63541927 | 58 | #include "props.h" |
3b02a68a | 59 | #include "sysfs.h" |
fcebe456 | 60 | #include "qgroup.h" |
1ec9a1ae | 61 | #include "tree-log.h" |
ebb8765b | 62 | #include "compression.h" |
f46b5a66 | 63 | |
abccd00f HM |
64 | #ifdef CONFIG_64BIT |
65 | /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI | |
66 | * structures are incorrect, as the timespec structure from userspace | |
67 | * is 4 bytes too small. We define these alternatives here to teach | |
68 | * the kernel about the 32-bit struct packing. | |
69 | */ | |
70 | struct btrfs_ioctl_timespec_32 { | |
71 | __u64 sec; | |
72 | __u32 nsec; | |
73 | } __attribute__ ((__packed__)); | |
74 | ||
75 | struct btrfs_ioctl_received_subvol_args_32 { | |
76 | char uuid[BTRFS_UUID_SIZE]; /* in */ | |
77 | __u64 stransid; /* in */ | |
78 | __u64 rtransid; /* out */ | |
79 | struct btrfs_ioctl_timespec_32 stime; /* in */ | |
80 | struct btrfs_ioctl_timespec_32 rtime; /* out */ | |
81 | __u64 flags; /* in */ | |
82 | __u64 reserved[16]; /* in */ | |
83 | } __attribute__ ((__packed__)); | |
84 | ||
85 | #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ | |
86 | struct btrfs_ioctl_received_subvol_args_32) | |
87 | #endif | |
88 | ||
89 | ||
416161db | 90 | static int btrfs_clone(struct inode *src, struct inode *inode, |
1c919a5e MF |
91 | u64 off, u64 olen, u64 olen_aligned, u64 destoff, |
92 | int no_time_update); | |
416161db | 93 | |
6cbff00f CH |
94 | /* Mask out flags that are inappropriate for the given type of inode. */ |
95 | static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags) | |
96 | { | |
97 | if (S_ISDIR(mode)) | |
98 | return flags; | |
99 | else if (S_ISREG(mode)) | |
100 | return flags & ~FS_DIRSYNC_FL; | |
101 | else | |
102 | return flags & (FS_NODUMP_FL | FS_NOATIME_FL); | |
103 | } | |
104 | ||
105 | /* | |
106 | * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl. | |
107 | */ | |
108 | static unsigned int btrfs_flags_to_ioctl(unsigned int flags) | |
109 | { | |
110 | unsigned int iflags = 0; | |
111 | ||
112 | if (flags & BTRFS_INODE_SYNC) | |
113 | iflags |= FS_SYNC_FL; | |
114 | if (flags & BTRFS_INODE_IMMUTABLE) | |
115 | iflags |= FS_IMMUTABLE_FL; | |
116 | if (flags & BTRFS_INODE_APPEND) | |
117 | iflags |= FS_APPEND_FL; | |
118 | if (flags & BTRFS_INODE_NODUMP) | |
119 | iflags |= FS_NODUMP_FL; | |
120 | if (flags & BTRFS_INODE_NOATIME) | |
121 | iflags |= FS_NOATIME_FL; | |
122 | if (flags & BTRFS_INODE_DIRSYNC) | |
123 | iflags |= FS_DIRSYNC_FL; | |
d0092bdd LZ |
124 | if (flags & BTRFS_INODE_NODATACOW) |
125 | iflags |= FS_NOCOW_FL; | |
126 | ||
13f48dc9 | 127 | if (flags & BTRFS_INODE_NOCOMPRESS) |
d0092bdd | 128 | iflags |= FS_NOCOMP_FL; |
13f48dc9 ST |
129 | else if (flags & BTRFS_INODE_COMPRESS) |
130 | iflags |= FS_COMPR_FL; | |
6cbff00f CH |
131 | |
132 | return iflags; | |
133 | } | |
134 | ||
135 | /* | |
136 | * Update inode->i_flags based on the btrfs internal flags. | |
137 | */ | |
138 | void btrfs_update_iflags(struct inode *inode) | |
139 | { | |
140 | struct btrfs_inode *ip = BTRFS_I(inode); | |
3cc79392 | 141 | unsigned int new_fl = 0; |
6cbff00f CH |
142 | |
143 | if (ip->flags & BTRFS_INODE_SYNC) | |
3cc79392 | 144 | new_fl |= S_SYNC; |
6cbff00f | 145 | if (ip->flags & BTRFS_INODE_IMMUTABLE) |
3cc79392 | 146 | new_fl |= S_IMMUTABLE; |
6cbff00f | 147 | if (ip->flags & BTRFS_INODE_APPEND) |
3cc79392 | 148 | new_fl |= S_APPEND; |
6cbff00f | 149 | if (ip->flags & BTRFS_INODE_NOATIME) |
3cc79392 | 150 | new_fl |= S_NOATIME; |
6cbff00f | 151 | if (ip->flags & BTRFS_INODE_DIRSYNC) |
3cc79392 FM |
152 | new_fl |= S_DIRSYNC; |
153 | ||
154 | set_mask_bits(&inode->i_flags, | |
155 | S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC, | |
156 | new_fl); | |
6cbff00f CH |
157 | } |
158 | ||
159 | /* | |
160 | * Inherit flags from the parent inode. | |
161 | * | |
e27425d6 | 162 | * Currently only the compression flags and the cow flags are inherited. |
6cbff00f CH |
163 | */ |
164 | void btrfs_inherit_iflags(struct inode *inode, struct inode *dir) | |
165 | { | |
0b4dcea5 CM |
166 | unsigned int flags; |
167 | ||
168 | if (!dir) | |
169 | return; | |
170 | ||
171 | flags = BTRFS_I(dir)->flags; | |
6cbff00f | 172 | |
e27425d6 JB |
173 | if (flags & BTRFS_INODE_NOCOMPRESS) { |
174 | BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS; | |
175 | BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS; | |
176 | } else if (flags & BTRFS_INODE_COMPRESS) { | |
177 | BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS; | |
178 | BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS; | |
179 | } | |
180 | ||
213490b3 | 181 | if (flags & BTRFS_INODE_NODATACOW) { |
e27425d6 | 182 | BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW; |
213490b3 LB |
183 | if (S_ISREG(inode->i_mode)) |
184 | BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM; | |
185 | } | |
6cbff00f | 186 | |
6cbff00f CH |
187 | btrfs_update_iflags(inode); |
188 | } | |
189 | ||
190 | static int btrfs_ioctl_getflags(struct file *file, void __user *arg) | |
191 | { | |
496ad9aa | 192 | struct btrfs_inode *ip = BTRFS_I(file_inode(file)); |
6cbff00f CH |
193 | unsigned int flags = btrfs_flags_to_ioctl(ip->flags); |
194 | ||
195 | if (copy_to_user(arg, &flags, sizeof(flags))) | |
196 | return -EFAULT; | |
197 | return 0; | |
198 | } | |
199 | ||
75e7cb7f LB |
200 | static int check_flags(unsigned int flags) |
201 | { | |
202 | if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ | |
203 | FS_NOATIME_FL | FS_NODUMP_FL | \ | |
204 | FS_SYNC_FL | FS_DIRSYNC_FL | \ | |
e1e8fb6a LZ |
205 | FS_NOCOMP_FL | FS_COMPR_FL | |
206 | FS_NOCOW_FL)) | |
75e7cb7f LB |
207 | return -EOPNOTSUPP; |
208 | ||
209 | if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) | |
210 | return -EINVAL; | |
211 | ||
75e7cb7f LB |
212 | return 0; |
213 | } | |
214 | ||
6cbff00f CH |
215 | static int btrfs_ioctl_setflags(struct file *file, void __user *arg) |
216 | { | |
496ad9aa | 217 | struct inode *inode = file_inode(file); |
0b246afa | 218 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
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 | /* | |
01327610 | 299 | * Revert back under same assumptions as above |
7e97b8da DS |
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 | 327 | |
0b246afa | 328 | if (fs_info->compress_type == BTRFS_COMPRESS_LZO) |
63541927 FDBM |
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); |
c2050a45 | 352 | inode->i_ctime = current_time(inode); |
6cbff00f | 353 | ret = btrfs_update_inode(trans, root, inode); |
6cbff00f | 354 | |
3a45bb20 | 355 | btrfs_end_transaction(trans); |
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 | { | |
0b246afa JM |
377 | struct inode *inode = file_inode(file); |
378 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
f7039b1d LD |
379 | struct btrfs_device *device; |
380 | struct request_queue *q; | |
381 | struct fstrim_range range; | |
382 | u64 minlen = ULLONG_MAX; | |
383 | u64 num_devices = 0; | |
815745cf | 384 | u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy); |
f7039b1d LD |
385 | int ret; |
386 | ||
387 | if (!capable(CAP_SYS_ADMIN)) | |
388 | return -EPERM; | |
389 | ||
1f78160c XG |
390 | rcu_read_lock(); |
391 | list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, | |
392 | dev_list) { | |
f7039b1d LD |
393 | if (!device->bdev) |
394 | continue; | |
395 | q = bdev_get_queue(device->bdev); | |
396 | if (blk_queue_discard(q)) { | |
397 | num_devices++; | |
50d0446e | 398 | minlen = min_t(u64, q->limits.discard_granularity, |
f7039b1d LD |
399 | minlen); |
400 | } | |
401 | } | |
1f78160c | 402 | rcu_read_unlock(); |
f4c697e6 | 403 | |
f7039b1d LD |
404 | if (!num_devices) |
405 | return -EOPNOTSUPP; | |
f7039b1d LD |
406 | if (copy_from_user(&range, arg, sizeof(range))) |
407 | return -EFAULT; | |
e515c18b LC |
408 | if (range.start > total_bytes || |
409 | range.len < fs_info->sb->s_blocksize) | |
f4c697e6 | 410 | return -EINVAL; |
f7039b1d | 411 | |
f4c697e6 | 412 | range.len = min(range.len, total_bytes - range.start); |
f7039b1d | 413 | range.minlen = max(range.minlen, minlen); |
2ff7e61e | 414 | ret = btrfs_trim_fs(fs_info, &range); |
f7039b1d LD |
415 | if (ret < 0) |
416 | return ret; | |
417 | ||
418 | if (copy_to_user(arg, &range, sizeof(range))) | |
419 | return -EFAULT; | |
420 | ||
421 | return 0; | |
422 | } | |
423 | ||
dd5f9615 SB |
424 | int btrfs_is_empty_uuid(u8 *uuid) |
425 | { | |
46e0f66a CM |
426 | int i; |
427 | ||
428 | for (i = 0; i < BTRFS_UUID_SIZE; i++) { | |
429 | if (uuid[i]) | |
430 | return 0; | |
431 | } | |
432 | return 1; | |
dd5f9615 SB |
433 | } |
434 | ||
d5c12070 | 435 | static noinline int create_subvol(struct inode *dir, |
cb8e7090 | 436 | struct dentry *dentry, |
72fd032e | 437 | char *name, int namelen, |
6f72c7e2 | 438 | u64 *async_transid, |
8696c533 | 439 | struct btrfs_qgroup_inherit *inherit) |
f46b5a66 | 440 | { |
0b246afa | 441 | struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); |
f46b5a66 CH |
442 | struct btrfs_trans_handle *trans; |
443 | struct btrfs_key key; | |
49a3c4d9 | 444 | struct btrfs_root_item *root_item; |
f46b5a66 CH |
445 | struct btrfs_inode_item *inode_item; |
446 | struct extent_buffer *leaf; | |
d5c12070 | 447 | struct btrfs_root *root = BTRFS_I(dir)->root; |
76dda93c | 448 | struct btrfs_root *new_root; |
d5c12070 | 449 | struct btrfs_block_rsv block_rsv; |
c2050a45 | 450 | struct timespec cur_time = current_time(dir); |
5662344b | 451 | struct inode *inode; |
f46b5a66 CH |
452 | int ret; |
453 | int err; | |
454 | u64 objectid; | |
455 | u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; | |
3de4586c | 456 | u64 index = 0; |
d5c12070 | 457 | u64 qgroup_reserved; |
8ea05e3a | 458 | uuid_le new_uuid; |
f46b5a66 | 459 | |
49a3c4d9 DS |
460 | root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); |
461 | if (!root_item) | |
462 | return -ENOMEM; | |
463 | ||
0b246afa | 464 | ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid); |
2fbe8c8a | 465 | if (ret) |
49a3c4d9 | 466 | goto fail_free; |
6a912213 | 467 | |
e09fe2d2 QW |
468 | /* |
469 | * Don't create subvolume whose level is not zero. Or qgroup will be | |
01327610 | 470 | * screwed up since it assumes subvolume qgroup's level to be 0. |
e09fe2d2 | 471 | */ |
49a3c4d9 DS |
472 | if (btrfs_qgroup_level(objectid)) { |
473 | ret = -ENOSPC; | |
474 | goto fail_free; | |
475 | } | |
e09fe2d2 | 476 | |
d5c12070 | 477 | btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); |
9ed74f2d | 478 | /* |
d5c12070 MX |
479 | * The same as the snapshot creation, please see the comment |
480 | * of create_snapshot(). | |
9ed74f2d | 481 | */ |
d5c12070 | 482 | ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, |
dd5f9615 | 483 | 8, &qgroup_reserved, false); |
d5c12070 | 484 | if (ret) |
49a3c4d9 | 485 | goto fail_free; |
d5c12070 MX |
486 | |
487 | trans = btrfs_start_transaction(root, 0); | |
488 | if (IS_ERR(trans)) { | |
489 | ret = PTR_ERR(trans); | |
2ff7e61e | 490 | btrfs_subvolume_release_metadata(fs_info, &block_rsv, |
de6e8200 | 491 | qgroup_reserved); |
49a3c4d9 | 492 | goto fail_free; |
d5c12070 MX |
493 | } |
494 | trans->block_rsv = &block_rsv; | |
495 | trans->bytes_reserved = block_rsv.size; | |
f46b5a66 | 496 | |
0b246afa | 497 | ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit); |
6f72c7e2 AJ |
498 | if (ret) |
499 | goto fail; | |
500 | ||
4d75f8a9 | 501 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); |
8e8a1e31 JB |
502 | if (IS_ERR(leaf)) { |
503 | ret = PTR_ERR(leaf); | |
504 | goto fail; | |
505 | } | |
f46b5a66 | 506 | |
b159fa28 | 507 | memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header)); |
f46b5a66 CH |
508 | btrfs_set_header_bytenr(leaf, leaf->start); |
509 | btrfs_set_header_generation(leaf, trans->transid); | |
5d4f98a2 | 510 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); |
f46b5a66 CH |
511 | btrfs_set_header_owner(leaf, objectid); |
512 | ||
0b246afa JM |
513 | write_extent_buffer_fsid(leaf, fs_info->fsid); |
514 | write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid); | |
f46b5a66 CH |
515 | btrfs_mark_buffer_dirty(leaf); |
516 | ||
49a3c4d9 | 517 | inode_item = &root_item->inode; |
3cae210f QW |
518 | btrfs_set_stack_inode_generation(inode_item, 1); |
519 | btrfs_set_stack_inode_size(inode_item, 3); | |
520 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 521 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 522 | fs_info->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; |
0b246afa | 555 | ret = btrfs_insert_root(trans, fs_info->tree_root, &key, |
49a3c4d9 | 556 | root_item); |
f46b5a66 CH |
557 | if (ret) |
558 | goto fail; | |
559 | ||
76dda93c | 560 | key.offset = (u64)-1; |
0b246afa | 561 | new_root = btrfs_read_fs_root_no_name(fs_info, &key); |
79787eaa | 562 | if (IS_ERR(new_root)) { |
79787eaa | 563 | ret = PTR_ERR(new_root); |
66642832 | 564 | btrfs_abort_transaction(trans, 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 */ | |
66642832 | 573 | btrfs_abort_transaction(trans, 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 | 585 | if (ret) { |
66642832 | 586 | btrfs_abort_transaction(trans, ret); |
79787eaa JM |
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 | 593 | if (ret) { |
66642832 | 594 | btrfs_abort_transaction(trans, 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 | ||
0b246afa | 602 | ret = btrfs_add_root_ref(trans, fs_info, |
4df27c4d | 603 | objectid, root->root_key.objectid, |
33345d01 | 604 | btrfs_ino(dir), index, name, namelen); |
76dda93c | 605 | BUG_ON(ret); |
f46b5a66 | 606 | |
0b246afa | 607 | ret = btrfs_uuid_tree_add(trans, fs_info, root_item->uuid, |
6bccf3ab | 608 | BTRFS_UUID_KEY_SUBVOL, objectid); |
dd5f9615 | 609 | if (ret) |
66642832 | 610 | btrfs_abort_transaction(trans, ret); |
dd5f9615 | 611 | |
f46b5a66 | 612 | fail: |
49a3c4d9 | 613 | kfree(root_item); |
d5c12070 MX |
614 | trans->block_rsv = NULL; |
615 | trans->bytes_reserved = 0; | |
2ff7e61e | 616 | btrfs_subvolume_release_metadata(fs_info, &block_rsv, qgroup_reserved); |
de6e8200 | 617 | |
72fd032e SW |
618 | if (async_transid) { |
619 | *async_transid = trans->transid; | |
3a45bb20 | 620 | err = btrfs_commit_transaction_async(trans, 1); |
00d71c9c | 621 | if (err) |
3a45bb20 | 622 | err = btrfs_commit_transaction(trans); |
72fd032e | 623 | } else { |
3a45bb20 | 624 | err = btrfs_commit_transaction(trans); |
72fd032e | 625 | } |
f46b5a66 CH |
626 | if (err && !ret) |
627 | ret = err; | |
1a65e24b | 628 | |
5662344b TI |
629 | if (!ret) { |
630 | inode = btrfs_lookup_dentry(dir, dentry); | |
de6e8200 LB |
631 | if (IS_ERR(inode)) |
632 | return PTR_ERR(inode); | |
5662344b TI |
633 | d_instantiate(dentry, inode); |
634 | } | |
f46b5a66 | 635 | return ret; |
49a3c4d9 DS |
636 | |
637 | fail_free: | |
638 | kfree(root_item); | |
639 | return ret; | |
f46b5a66 CH |
640 | } |
641 | ||
9ea24bbe | 642 | static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root) |
8257b2dc MX |
643 | { |
644 | s64 writers; | |
645 | DEFINE_WAIT(wait); | |
646 | ||
647 | do { | |
648 | prepare_to_wait(&root->subv_writers->wait, &wait, | |
649 | TASK_UNINTERRUPTIBLE); | |
650 | ||
651 | writers = percpu_counter_sum(&root->subv_writers->counter); | |
652 | if (writers) | |
653 | schedule(); | |
654 | ||
655 | finish_wait(&root->subv_writers->wait, &wait); | |
656 | } while (writers); | |
657 | } | |
658 | ||
e9662f70 MX |
659 | static int create_snapshot(struct btrfs_root *root, struct inode *dir, |
660 | struct dentry *dentry, char *name, int namelen, | |
661 | u64 *async_transid, bool readonly, | |
662 | struct btrfs_qgroup_inherit *inherit) | |
f46b5a66 | 663 | { |
0b246afa | 664 | struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); |
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 | |
578def7c | 693 | btrfs_wait_ordered_extents(root, -1, 0, (u64)-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 | ||
0b246afa | 724 | spin_lock(&fs_info->trans_lock); |
f46b5a66 CH |
725 | list_add(&pending_snapshot->list, |
726 | &trans->transaction->pending_snapshots); | |
0b246afa | 727 | spin_unlock(&fs_info->trans_lock); |
72fd032e SW |
728 | if (async_transid) { |
729 | *async_transid = trans->transid; | |
3a45bb20 | 730 | ret = btrfs_commit_transaction_async(trans, 1); |
00d71c9c | 731 | if (ret) |
3a45bb20 | 732 | ret = btrfs_commit_transaction(trans); |
72fd032e | 733 | } else { |
3a45bb20 | 734 | ret = btrfs_commit_transaction(trans); |
72fd032e | 735 | } |
aec8030a | 736 | if (ret) |
c37b2b62 | 737 | goto fail; |
a22285a6 YZ |
738 | |
739 | ret = pending_snapshot->error; | |
740 | if (ret) | |
741 | goto fail; | |
742 | ||
d3797308 CM |
743 | ret = btrfs_orphan_cleanup(pending_snapshot->snap); |
744 | if (ret) | |
745 | goto fail; | |
746 | ||
2b0143b5 | 747 | inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); |
2e4bfab9 YZ |
748 | if (IS_ERR(inode)) { |
749 | ret = PTR_ERR(inode); | |
750 | goto fail; | |
751 | } | |
5662344b | 752 | |
2e4bfab9 YZ |
753 | d_instantiate(dentry, inode); |
754 | ret = 0; | |
755 | fail: | |
2ff7e61e | 756 | btrfs_subvolume_release_metadata(fs_info, |
d5c12070 MX |
757 | &pending_snapshot->block_rsv, |
758 | pending_snapshot->qgroup_reserved); | |
a1ee7362 | 759 | dec_and_free: |
9ea24bbe FM |
760 | if (atomic_dec_and_test(&root->will_be_snapshoted)) |
761 | wake_up_atomic_t(&root->will_be_snapshoted); | |
b0c0ea63 DS |
762 | free_pending: |
763 | kfree(pending_snapshot->root_item); | |
8546b570 | 764 | btrfs_free_path(pending_snapshot->path); |
a1ee7362 DS |
765 | kfree(pending_snapshot); |
766 | ||
f46b5a66 CH |
767 | return ret; |
768 | } | |
769 | ||
4260f7c7 SW |
770 | /* copy of may_delete in fs/namei.c() |
771 | * Check whether we can remove a link victim from directory dir, check | |
772 | * whether the type of victim is right. | |
773 | * 1. We can't do it if dir is read-only (done in permission()) | |
774 | * 2. We should have write and exec permissions on dir | |
775 | * 3. We can't remove anything from append-only dir | |
776 | * 4. We can't do anything with immutable dir (done in permission()) | |
777 | * 5. If the sticky bit on dir is set we should either | |
778 | * a. be owner of dir, or | |
779 | * b. be owner of victim, or | |
780 | * c. have CAP_FOWNER capability | |
01327610 | 781 | * 6. If the victim is append-only or immutable we can't do anything with |
4260f7c7 SW |
782 | * links pointing to it. |
783 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
784 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
785 | * 9. We can't remove a root or mountpoint. | |
786 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
787 | * nfs_async_unlink(). | |
788 | */ | |
789 | ||
67871254 | 790 | static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) |
4260f7c7 SW |
791 | { |
792 | int error; | |
793 | ||
2b0143b5 | 794 | if (d_really_is_negative(victim)) |
4260f7c7 SW |
795 | return -ENOENT; |
796 | ||
2b0143b5 | 797 | BUG_ON(d_inode(victim->d_parent) != dir); |
4fa6b5ec | 798 | audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); |
4260f7c7 SW |
799 | |
800 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); | |
801 | if (error) | |
802 | return error; | |
803 | if (IS_APPEND(dir)) | |
804 | return -EPERM; | |
2b0143b5 DH |
805 | if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || |
806 | IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) | |
4260f7c7 SW |
807 | return -EPERM; |
808 | if (isdir) { | |
e36cb0b8 | 809 | if (!d_is_dir(victim)) |
4260f7c7 SW |
810 | return -ENOTDIR; |
811 | if (IS_ROOT(victim)) | |
812 | return -EBUSY; | |
e36cb0b8 | 813 | } else if (d_is_dir(victim)) |
4260f7c7 SW |
814 | return -EISDIR; |
815 | if (IS_DEADDIR(dir)) | |
816 | return -ENOENT; | |
817 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
818 | return -EBUSY; | |
819 | return 0; | |
820 | } | |
821 | ||
cb8e7090 CH |
822 | /* copy of may_create in fs/namei.c() */ |
823 | static inline int btrfs_may_create(struct inode *dir, struct dentry *child) | |
824 | { | |
2b0143b5 | 825 | if (d_really_is_positive(child)) |
cb8e7090 CH |
826 | return -EEXIST; |
827 | if (IS_DEADDIR(dir)) | |
828 | return -ENOENT; | |
829 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); | |
830 | } | |
831 | ||
832 | /* | |
833 | * Create a new subvolume below @parent. This is largely modeled after | |
834 | * sys_mkdirat and vfs_mkdir, but we only do a single component lookup | |
835 | * inside this filesystem so it's quite a bit simpler. | |
836 | */ | |
92872094 | 837 | static noinline int btrfs_mksubvol(const struct path *parent, |
76dda93c | 838 | char *name, int namelen, |
72fd032e | 839 | struct btrfs_root *snap_src, |
6f72c7e2 | 840 | u64 *async_transid, bool readonly, |
8696c533 | 841 | struct btrfs_qgroup_inherit *inherit) |
cb8e7090 | 842 | { |
0b246afa JM |
843 | struct inode *dir = d_inode(parent->dentry); |
844 | struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); | |
cb8e7090 CH |
845 | struct dentry *dentry; |
846 | int error; | |
847 | ||
00235411 AV |
848 | error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); |
849 | if (error == -EINTR) | |
850 | return error; | |
cb8e7090 CH |
851 | |
852 | dentry = lookup_one_len(name, parent->dentry, namelen); | |
853 | error = PTR_ERR(dentry); | |
854 | if (IS_ERR(dentry)) | |
855 | goto out_unlock; | |
856 | ||
76dda93c | 857 | error = btrfs_may_create(dir, dentry); |
cb8e7090 | 858 | if (error) |
a874a63e | 859 | goto out_dput; |
cb8e7090 | 860 | |
9c52057c CM |
861 | /* |
862 | * even if this name doesn't exist, we may get hash collisions. | |
863 | * check for them now when we can safely fail | |
864 | */ | |
865 | error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, | |
866 | dir->i_ino, name, | |
867 | namelen); | |
868 | if (error) | |
869 | goto out_dput; | |
870 | ||
0b246afa | 871 | down_read(&fs_info->subvol_sem); |
76dda93c YZ |
872 | |
873 | if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) | |
874 | goto out_up_read; | |
875 | ||
3de4586c | 876 | if (snap_src) { |
e9662f70 | 877 | error = create_snapshot(snap_src, dir, dentry, name, namelen, |
6f72c7e2 | 878 | async_transid, readonly, inherit); |
3de4586c | 879 | } else { |
d5c12070 MX |
880 | error = create_subvol(dir, dentry, name, namelen, |
881 | async_transid, inherit); | |
3de4586c | 882 | } |
76dda93c YZ |
883 | if (!error) |
884 | fsnotify_mkdir(dir, dentry); | |
885 | out_up_read: | |
0b246afa | 886 | up_read(&fs_info->subvol_sem); |
cb8e7090 CH |
887 | out_dput: |
888 | dput(dentry); | |
889 | out_unlock: | |
5955102c | 890 | inode_unlock(dir); |
cb8e7090 CH |
891 | return error; |
892 | } | |
893 | ||
4cb5300b CM |
894 | /* |
895 | * When we're defragging a range, we don't want to kick it off again | |
896 | * if it is really just waiting for delalloc to send it down. | |
897 | * If we find a nice big extent or delalloc range for the bytes in the | |
898 | * file you want to defrag, we return 0 to let you know to skip this | |
899 | * part of the file | |
900 | */ | |
aab110ab | 901 | static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) |
4cb5300b CM |
902 | { |
903 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | |
904 | struct extent_map *em = NULL; | |
905 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
906 | u64 end; | |
907 | ||
908 | read_lock(&em_tree->lock); | |
09cbfeaf | 909 | em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); |
4cb5300b CM |
910 | read_unlock(&em_tree->lock); |
911 | ||
912 | if (em) { | |
913 | end = extent_map_end(em); | |
914 | free_extent_map(em); | |
915 | if (end - offset > thresh) | |
916 | return 0; | |
917 | } | |
918 | /* if we already have a nice delalloc here, just stop */ | |
919 | thresh /= 2; | |
920 | end = count_range_bits(io_tree, &offset, offset + thresh, | |
921 | thresh, EXTENT_DELALLOC, 1); | |
922 | if (end >= thresh) | |
923 | return 0; | |
924 | return 1; | |
925 | } | |
926 | ||
927 | /* | |
928 | * helper function to walk through a file and find extents | |
929 | * newer than a specific transid, and smaller than thresh. | |
930 | * | |
931 | * This is used by the defragging code to find new and small | |
932 | * extents | |
933 | */ | |
934 | static int find_new_extents(struct btrfs_root *root, | |
935 | struct inode *inode, u64 newer_than, | |
aab110ab | 936 | u64 *off, u32 thresh) |
4cb5300b CM |
937 | { |
938 | struct btrfs_path *path; | |
939 | struct btrfs_key min_key; | |
4cb5300b CM |
940 | struct extent_buffer *leaf; |
941 | struct btrfs_file_extent_item *extent; | |
942 | int type; | |
943 | int ret; | |
a4689d2b | 944 | u64 ino = btrfs_ino(inode); |
4cb5300b CM |
945 | |
946 | path = btrfs_alloc_path(); | |
947 | if (!path) | |
948 | return -ENOMEM; | |
949 | ||
a4689d2b | 950 | min_key.objectid = ino; |
4cb5300b CM |
951 | min_key.type = BTRFS_EXTENT_DATA_KEY; |
952 | min_key.offset = *off; | |
953 | ||
67871254 | 954 | while (1) { |
6174d3cb | 955 | ret = btrfs_search_forward(root, &min_key, path, newer_than); |
4cb5300b CM |
956 | if (ret != 0) |
957 | goto none; | |
f094c9bd | 958 | process_slot: |
a4689d2b | 959 | if (min_key.objectid != ino) |
4cb5300b CM |
960 | goto none; |
961 | if (min_key.type != BTRFS_EXTENT_DATA_KEY) | |
962 | goto none; | |
963 | ||
964 | leaf = path->nodes[0]; | |
965 | extent = btrfs_item_ptr(leaf, path->slots[0], | |
966 | struct btrfs_file_extent_item); | |
967 | ||
968 | type = btrfs_file_extent_type(leaf, extent); | |
969 | if (type == BTRFS_FILE_EXTENT_REG && | |
970 | btrfs_file_extent_num_bytes(leaf, extent) < thresh && | |
971 | check_defrag_in_cache(inode, min_key.offset, thresh)) { | |
972 | *off = min_key.offset; | |
973 | btrfs_free_path(path); | |
974 | return 0; | |
975 | } | |
976 | ||
f094c9bd FM |
977 | path->slots[0]++; |
978 | if (path->slots[0] < btrfs_header_nritems(leaf)) { | |
979 | btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); | |
980 | goto process_slot; | |
981 | } | |
982 | ||
4cb5300b CM |
983 | if (min_key.offset == (u64)-1) |
984 | goto none; | |
985 | ||
986 | min_key.offset++; | |
987 | btrfs_release_path(path); | |
988 | } | |
989 | none: | |
990 | btrfs_free_path(path); | |
991 | return -ENOENT; | |
992 | } | |
993 | ||
6c282eb4 | 994 | static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) |
17ce6ef8 LB |
995 | { |
996 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
6c282eb4 LZ |
997 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
998 | struct extent_map *em; | |
09cbfeaf | 999 | u64 len = PAGE_SIZE; |
17ce6ef8 | 1000 | |
6c282eb4 LZ |
1001 | /* |
1002 | * hopefully we have this extent in the tree already, try without | |
1003 | * the full extent lock | |
1004 | */ | |
17ce6ef8 | 1005 | read_lock(&em_tree->lock); |
6c282eb4 | 1006 | em = lookup_extent_mapping(em_tree, start, len); |
17ce6ef8 LB |
1007 | read_unlock(&em_tree->lock); |
1008 | ||
6c282eb4 | 1009 | if (!em) { |
308d9800 FM |
1010 | struct extent_state *cached = NULL; |
1011 | u64 end = start + len - 1; | |
1012 | ||
6c282eb4 | 1013 | /* get the big lock and read metadata off disk */ |
ff13db41 | 1014 | lock_extent_bits(io_tree, start, end, &cached); |
6c282eb4 | 1015 | em = btrfs_get_extent(inode, NULL, 0, start, len, 0); |
308d9800 | 1016 | unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS); |
6c282eb4 LZ |
1017 | |
1018 | if (IS_ERR(em)) | |
1019 | return NULL; | |
1020 | } | |
1021 | ||
1022 | return em; | |
1023 | } | |
17ce6ef8 | 1024 | |
6c282eb4 LZ |
1025 | static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) |
1026 | { | |
1027 | struct extent_map *next; | |
1028 | bool ret = true; | |
1029 | ||
1030 | /* this is the last extent */ | |
1031 | if (em->start + em->len >= i_size_read(inode)) | |
1032 | return false; | |
1033 | ||
1034 | next = defrag_lookup_extent(inode, em->start + em->len); | |
e9512d72 CM |
1035 | if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) |
1036 | ret = false; | |
1037 | else if ((em->block_start + em->block_len == next->block_start) && | |
ee22184b | 1038 | (em->block_len > SZ_128K && next->block_len > SZ_128K)) |
6c282eb4 LZ |
1039 | ret = false; |
1040 | ||
1041 | free_extent_map(next); | |
17ce6ef8 LB |
1042 | return ret; |
1043 | } | |
1044 | ||
aab110ab | 1045 | static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, |
a43a2111 AM |
1046 | u64 *last_len, u64 *skip, u64 *defrag_end, |
1047 | int compress) | |
940100a4 | 1048 | { |
6c282eb4 | 1049 | struct extent_map *em; |
940100a4 | 1050 | int ret = 1; |
6c282eb4 | 1051 | bool next_mergeable = true; |
4a3560c4 | 1052 | bool prev_mergeable = true; |
940100a4 CM |
1053 | |
1054 | /* | |
008873ea | 1055 | * make sure that once we start defragging an extent, we keep on |
940100a4 CM |
1056 | * defragging it |
1057 | */ | |
1058 | if (start < *defrag_end) | |
1059 | return 1; | |
1060 | ||
1061 | *skip = 0; | |
1062 | ||
6c282eb4 LZ |
1063 | em = defrag_lookup_extent(inode, start); |
1064 | if (!em) | |
1065 | return 0; | |
940100a4 CM |
1066 | |
1067 | /* this will cover holes, and inline extents */ | |
17ce6ef8 | 1068 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) { |
940100a4 | 1069 | ret = 0; |
17ce6ef8 LB |
1070 | goto out; |
1071 | } | |
1072 | ||
4a3560c4 LB |
1073 | if (!*defrag_end) |
1074 | prev_mergeable = false; | |
1075 | ||
6c282eb4 | 1076 | next_mergeable = defrag_check_next_extent(inode, em); |
940100a4 | 1077 | /* |
6c282eb4 LZ |
1078 | * we hit a real extent, if it is big or the next extent is not a |
1079 | * real extent, don't bother defragging it | |
940100a4 | 1080 | */ |
a43a2111 | 1081 | if (!compress && (*last_len == 0 || *last_len >= thresh) && |
4a3560c4 | 1082 | (em->len >= thresh || (!next_mergeable && !prev_mergeable))) |
940100a4 | 1083 | ret = 0; |
17ce6ef8 | 1084 | out: |
940100a4 CM |
1085 | /* |
1086 | * last_len ends up being a counter of how many bytes we've defragged. | |
1087 | * every time we choose not to defrag an extent, we reset *last_len | |
1088 | * so that the next tiny extent will force a defrag. | |
1089 | * | |
1090 | * The end result of this is that tiny extents before a single big | |
1091 | * extent will force at least part of that big extent to be defragged. | |
1092 | */ | |
1093 | if (ret) { | |
940100a4 CM |
1094 | *defrag_end = extent_map_end(em); |
1095 | } else { | |
1096 | *last_len = 0; | |
1097 | *skip = extent_map_end(em); | |
1098 | *defrag_end = 0; | |
1099 | } | |
1100 | ||
1101 | free_extent_map(em); | |
1102 | return ret; | |
1103 | } | |
1104 | ||
4cb5300b CM |
1105 | /* |
1106 | * it doesn't do much good to defrag one or two pages | |
1107 | * at a time. This pulls in a nice chunk of pages | |
1108 | * to COW and defrag. | |
1109 | * | |
1110 | * It also makes sure the delalloc code has enough | |
1111 | * dirty data to avoid making new small extents as part | |
1112 | * of the defrag | |
1113 | * | |
1114 | * It's a good idea to start RA on this range | |
1115 | * before calling this. | |
1116 | */ | |
1117 | static int cluster_pages_for_defrag(struct inode *inode, | |
1118 | struct page **pages, | |
1119 | unsigned long start_index, | |
c41570c9 | 1120 | unsigned long num_pages) |
f46b5a66 | 1121 | { |
4cb5300b CM |
1122 | unsigned long file_end; |
1123 | u64 isize = i_size_read(inode); | |
1124 | u64 page_start; | |
1125 | u64 page_end; | |
1f12bd06 | 1126 | u64 page_cnt; |
4cb5300b CM |
1127 | int ret; |
1128 | int i; | |
1129 | int i_done; | |
3eaa2885 | 1130 | struct btrfs_ordered_extent *ordered; |
4cb5300b | 1131 | struct extent_state *cached_state = NULL; |
600a45e1 | 1132 | struct extent_io_tree *tree; |
3b16a4e3 | 1133 | gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); |
4cb5300b | 1134 | |
09cbfeaf | 1135 | file_end = (isize - 1) >> PAGE_SHIFT; |
1f12bd06 LB |
1136 | if (!isize || start_index > file_end) |
1137 | return 0; | |
1138 | ||
1139 | page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); | |
4cb5300b CM |
1140 | |
1141 | ret = btrfs_delalloc_reserve_space(inode, | |
09cbfeaf KS |
1142 | start_index << PAGE_SHIFT, |
1143 | page_cnt << PAGE_SHIFT); | |
4cb5300b CM |
1144 | if (ret) |
1145 | return ret; | |
4cb5300b | 1146 | i_done = 0; |
600a45e1 | 1147 | tree = &BTRFS_I(inode)->io_tree; |
4cb5300b CM |
1148 | |
1149 | /* step one, lock all the pages */ | |
1f12bd06 | 1150 | for (i = 0; i < page_cnt; i++) { |
4cb5300b | 1151 | struct page *page; |
600a45e1 | 1152 | again: |
a94733d0 | 1153 | page = find_or_create_page(inode->i_mapping, |
600a45e1 | 1154 | start_index + i, mask); |
4cb5300b CM |
1155 | if (!page) |
1156 | break; | |
1157 | ||
600a45e1 | 1158 | page_start = page_offset(page); |
09cbfeaf | 1159 | page_end = page_start + PAGE_SIZE - 1; |
600a45e1 | 1160 | while (1) { |
308d9800 | 1161 | lock_extent_bits(tree, page_start, page_end, |
ff13db41 | 1162 | &cached_state); |
600a45e1 MX |
1163 | ordered = btrfs_lookup_ordered_extent(inode, |
1164 | page_start); | |
308d9800 FM |
1165 | unlock_extent_cached(tree, page_start, page_end, |
1166 | &cached_state, GFP_NOFS); | |
600a45e1 MX |
1167 | if (!ordered) |
1168 | break; | |
1169 | ||
1170 | unlock_page(page); | |
1171 | btrfs_start_ordered_extent(inode, ordered, 1); | |
1172 | btrfs_put_ordered_extent(ordered); | |
1173 | lock_page(page); | |
1f12bd06 LB |
1174 | /* |
1175 | * we unlocked the page above, so we need check if | |
1176 | * it was released or not. | |
1177 | */ | |
1178 | if (page->mapping != inode->i_mapping) { | |
1179 | unlock_page(page); | |
09cbfeaf | 1180 | put_page(page); |
1f12bd06 LB |
1181 | goto again; |
1182 | } | |
600a45e1 MX |
1183 | } |
1184 | ||
4cb5300b CM |
1185 | if (!PageUptodate(page)) { |
1186 | btrfs_readpage(NULL, page); | |
1187 | lock_page(page); | |
1188 | if (!PageUptodate(page)) { | |
1189 | unlock_page(page); | |
09cbfeaf | 1190 | put_page(page); |
4cb5300b CM |
1191 | ret = -EIO; |
1192 | break; | |
1193 | } | |
1194 | } | |
600a45e1 | 1195 | |
600a45e1 MX |
1196 | if (page->mapping != inode->i_mapping) { |
1197 | unlock_page(page); | |
09cbfeaf | 1198 | put_page(page); |
600a45e1 MX |
1199 | goto again; |
1200 | } | |
1201 | ||
4cb5300b CM |
1202 | pages[i] = page; |
1203 | i_done++; | |
1204 | } | |
1205 | if (!i_done || ret) | |
1206 | goto out; | |
1207 | ||
1208 | if (!(inode->i_sb->s_flags & MS_ACTIVE)) | |
1209 | goto out; | |
1210 | ||
1211 | /* | |
1212 | * so now we have a nice long stream of locked | |
1213 | * and up to date pages, lets wait on them | |
1214 | */ | |
1215 | for (i = 0; i < i_done; i++) | |
1216 | wait_on_page_writeback(pages[i]); | |
1217 | ||
1218 | page_start = page_offset(pages[0]); | |
09cbfeaf | 1219 | page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; |
4cb5300b CM |
1220 | |
1221 | lock_extent_bits(&BTRFS_I(inode)->io_tree, | |
ff13db41 | 1222 | page_start, page_end - 1, &cached_state); |
4cb5300b CM |
1223 | clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, |
1224 | page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | | |
9e8a4a8b LB |
1225 | EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, |
1226 | &cached_state, GFP_NOFS); | |
4cb5300b | 1227 | |
1f12bd06 | 1228 | if (i_done != page_cnt) { |
9e0baf60 JB |
1229 | spin_lock(&BTRFS_I(inode)->lock); |
1230 | BTRFS_I(inode)->outstanding_extents++; | |
1231 | spin_unlock(&BTRFS_I(inode)->lock); | |
4cb5300b | 1232 | btrfs_delalloc_release_space(inode, |
09cbfeaf KS |
1233 | start_index << PAGE_SHIFT, |
1234 | (page_cnt - i_done) << PAGE_SHIFT); | |
4cb5300b CM |
1235 | } |
1236 | ||
1237 | ||
9e8a4a8b | 1238 | set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, |
018ed4f7 | 1239 | &cached_state); |
4cb5300b CM |
1240 | |
1241 | unlock_extent_cached(&BTRFS_I(inode)->io_tree, | |
1242 | page_start, page_end - 1, &cached_state, | |
1243 | GFP_NOFS); | |
1244 | ||
1245 | for (i = 0; i < i_done; i++) { | |
1246 | clear_page_dirty_for_io(pages[i]); | |
1247 | ClearPageChecked(pages[i]); | |
1248 | set_page_extent_mapped(pages[i]); | |
1249 | set_page_dirty(pages[i]); | |
1250 | unlock_page(pages[i]); | |
09cbfeaf | 1251 | put_page(pages[i]); |
4cb5300b CM |
1252 | } |
1253 | return i_done; | |
1254 | out: | |
1255 | for (i = 0; i < i_done; i++) { | |
1256 | unlock_page(pages[i]); | |
09cbfeaf | 1257 | put_page(pages[i]); |
4cb5300b | 1258 | } |
7cf5b976 | 1259 | btrfs_delalloc_release_space(inode, |
09cbfeaf KS |
1260 | start_index << PAGE_SHIFT, |
1261 | page_cnt << PAGE_SHIFT); | |
4cb5300b CM |
1262 | return ret; |
1263 | ||
1264 | } | |
1265 | ||
1266 | int btrfs_defrag_file(struct inode *inode, struct file *file, | |
1267 | struct btrfs_ioctl_defrag_range_args *range, | |
1268 | u64 newer_than, unsigned long max_to_defrag) | |
1269 | { | |
0b246afa | 1270 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
4cb5300b | 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 | ||
0b246afa JM |
1368 | if (btrfs_defrag_cancelled(fs_info)) { |
1369 | btrfs_debug(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 | */ | |
0b246afa JM |
1457 | atomic_inc(&fs_info->async_submit_draining); |
1458 | while (atomic_read(&fs_info->nr_async_submits) || | |
1459 | atomic_read(&fs_info->async_delalloc_pages)) { | |
1460 | wait_event(fs_info->async_submit_wait, | |
1461 | (atomic_read(&fs_info->nr_async_submits) == 0 && | |
1462 | atomic_read(&fs_info->async_delalloc_pages) == 0)); | |
1e701a32 | 1463 | } |
0b246afa | 1464 | atomic_dec(&fs_info->async_submit_draining); |
1e701a32 CM |
1465 | } |
1466 | ||
1a419d85 | 1467 | if (range->compress_type == BTRFS_COMPRESS_LZO) { |
0b246afa | 1468 | btrfs_set_fs_incompat(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 | 1487 | { |
0b246afa JM |
1488 | struct inode *inode = file_inode(file); |
1489 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
f46b5a66 CH |
1490 | u64 new_size; |
1491 | u64 old_size; | |
1492 | u64 devid = 1; | |
0b246afa | 1493 | struct btrfs_root *root = BTRFS_I(inode)->root; |
f46b5a66 CH |
1494 | struct btrfs_ioctl_vol_args *vol_args; |
1495 | struct btrfs_trans_handle *trans; | |
1496 | struct btrfs_device *device = NULL; | |
1497 | char *sizestr; | |
9a40f122 | 1498 | char *retptr; |
f46b5a66 CH |
1499 | char *devstr = NULL; |
1500 | int ret = 0; | |
f46b5a66 CH |
1501 | int mod = 0; |
1502 | ||
e441d54d CM |
1503 | if (!capable(CAP_SYS_ADMIN)) |
1504 | return -EPERM; | |
1505 | ||
198605a8 MX |
1506 | ret = mnt_want_write_file(file); |
1507 | if (ret) | |
1508 | return ret; | |
1509 | ||
0b246afa | 1510 | if (atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { |
97547676 | 1511 | mnt_drop_write_file(file); |
e57138b3 | 1512 | return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
c9e9f97b ID |
1513 | } |
1514 | ||
0b246afa | 1515 | mutex_lock(&fs_info->volume_mutex); |
dae7b665 | 1516 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
1517 | if (IS_ERR(vol_args)) { |
1518 | ret = PTR_ERR(vol_args); | |
1519 | goto out; | |
1520 | } | |
5516e595 MF |
1521 | |
1522 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
f46b5a66 | 1523 | |
f46b5a66 CH |
1524 | sizestr = vol_args->name; |
1525 | devstr = strchr(sizestr, ':'); | |
1526 | if (devstr) { | |
f46b5a66 CH |
1527 | sizestr = devstr + 1; |
1528 | *devstr = '\0'; | |
1529 | devstr = vol_args->name; | |
58dfae63 Z |
1530 | ret = kstrtoull(devstr, 10, &devid); |
1531 | if (ret) | |
1532 | goto out_free; | |
dfd79829 MX |
1533 | if (!devid) { |
1534 | ret = -EINVAL; | |
1535 | goto out_free; | |
1536 | } | |
0b246afa | 1537 | btrfs_info(fs_info, "resizing devid %llu", devid); |
f46b5a66 | 1538 | } |
dba60f3f | 1539 | |
0b246afa | 1540 | device = btrfs_find_device(fs_info, devid, NULL, NULL); |
f46b5a66 | 1541 | if (!device) { |
0b246afa JM |
1542 | btrfs_info(fs_info, "resizer unable to find device %llu", |
1543 | devid); | |
dfd79829 | 1544 | ret = -ENODEV; |
c9e9f97b | 1545 | goto out_free; |
f46b5a66 | 1546 | } |
dba60f3f MX |
1547 | |
1548 | if (!device->writeable) { | |
0b246afa | 1549 | btrfs_info(fs_info, |
efe120a0 | 1550 | "resizer unable to apply on readonly device %llu", |
c1c9ff7c | 1551 | devid); |
dfd79829 | 1552 | ret = -EPERM; |
4e42ae1b LB |
1553 | goto out_free; |
1554 | } | |
1555 | ||
f46b5a66 CH |
1556 | if (!strcmp(sizestr, "max")) |
1557 | new_size = device->bdev->bd_inode->i_size; | |
1558 | else { | |
1559 | if (sizestr[0] == '-') { | |
1560 | mod = -1; | |
1561 | sizestr++; | |
1562 | } else if (sizestr[0] == '+') { | |
1563 | mod = 1; | |
1564 | sizestr++; | |
1565 | } | |
9a40f122 GH |
1566 | new_size = memparse(sizestr, &retptr); |
1567 | if (*retptr != '\0' || new_size == 0) { | |
f46b5a66 | 1568 | ret = -EINVAL; |
c9e9f97b | 1569 | goto out_free; |
f46b5a66 CH |
1570 | } |
1571 | } | |
1572 | ||
63a212ab | 1573 | if (device->is_tgtdev_for_dev_replace) { |
dfd79829 | 1574 | ret = -EPERM; |
63a212ab SB |
1575 | goto out_free; |
1576 | } | |
1577 | ||
7cc8e58d | 1578 | old_size = btrfs_device_get_total_bytes(device); |
f46b5a66 CH |
1579 | |
1580 | if (mod < 0) { | |
1581 | if (new_size > old_size) { | |
1582 | ret = -EINVAL; | |
c9e9f97b | 1583 | goto out_free; |
f46b5a66 CH |
1584 | } |
1585 | new_size = old_size - new_size; | |
1586 | } else if (mod > 0) { | |
eb8052e0 | 1587 | if (new_size > ULLONG_MAX - old_size) { |
902c68a4 | 1588 | ret = -ERANGE; |
eb8052e0 WF |
1589 | goto out_free; |
1590 | } | |
f46b5a66 CH |
1591 | new_size = old_size + new_size; |
1592 | } | |
1593 | ||
ee22184b | 1594 | if (new_size < SZ_256M) { |
f46b5a66 | 1595 | ret = -EINVAL; |
c9e9f97b | 1596 | goto out_free; |
f46b5a66 CH |
1597 | } |
1598 | if (new_size > device->bdev->bd_inode->i_size) { | |
1599 | ret = -EFBIG; | |
c9e9f97b | 1600 | goto out_free; |
f46b5a66 CH |
1601 | } |
1602 | ||
0b246afa JM |
1603 | new_size = div_u64(new_size, fs_info->sectorsize); |
1604 | new_size *= fs_info->sectorsize; | |
f46b5a66 | 1605 | |
0b246afa JM |
1606 | btrfs_info_in_rcu(fs_info, "new size for %s is %llu", |
1607 | rcu_str_deref(device->name), new_size); | |
f46b5a66 CH |
1608 | |
1609 | if (new_size > old_size) { | |
a22285a6 | 1610 | trans = btrfs_start_transaction(root, 0); |
98d5dc13 TI |
1611 | if (IS_ERR(trans)) { |
1612 | ret = PTR_ERR(trans); | |
c9e9f97b | 1613 | goto out_free; |
98d5dc13 | 1614 | } |
f46b5a66 | 1615 | ret = btrfs_grow_device(trans, device, new_size); |
3a45bb20 | 1616 | btrfs_commit_transaction(trans); |
ece7d20e | 1617 | } else if (new_size < old_size) { |
f46b5a66 | 1618 | ret = btrfs_shrink_device(device, new_size); |
0253f40e | 1619 | } /* equal, nothing need to do */ |
f46b5a66 | 1620 | |
c9e9f97b | 1621 | out_free: |
f46b5a66 | 1622 | kfree(vol_args); |
c9e9f97b | 1623 | out: |
0b246afa JM |
1624 | mutex_unlock(&fs_info->volume_mutex); |
1625 | atomic_set(&fs_info->mutually_exclusive_operation_running, 0); | |
18f39c41 | 1626 | mnt_drop_write_file(file); |
f46b5a66 CH |
1627 | return ret; |
1628 | } | |
1629 | ||
72fd032e | 1630 | static noinline int btrfs_ioctl_snap_create_transid(struct file *file, |
6f72c7e2 AJ |
1631 | char *name, unsigned long fd, int subvol, |
1632 | u64 *transid, bool readonly, | |
8696c533 | 1633 | struct btrfs_qgroup_inherit *inherit) |
f46b5a66 | 1634 | { |
f46b5a66 | 1635 | int namelen; |
3de4586c | 1636 | int ret = 0; |
f46b5a66 | 1637 | |
325c50e3 JM |
1638 | if (!S_ISDIR(file_inode(file)->i_mode)) |
1639 | return -ENOTDIR; | |
1640 | ||
a874a63e LB |
1641 | ret = mnt_want_write_file(file); |
1642 | if (ret) | |
1643 | goto out; | |
1644 | ||
72fd032e SW |
1645 | namelen = strlen(name); |
1646 | if (strchr(name, '/')) { | |
f46b5a66 | 1647 | ret = -EINVAL; |
a874a63e | 1648 | goto out_drop_write; |
f46b5a66 CH |
1649 | } |
1650 | ||
16780cab CM |
1651 | if (name[0] == '.' && |
1652 | (namelen == 1 || (name[1] == '.' && namelen == 2))) { | |
1653 | ret = -EEXIST; | |
a874a63e | 1654 | goto out_drop_write; |
16780cab CM |
1655 | } |
1656 | ||
3de4586c | 1657 | if (subvol) { |
72fd032e | 1658 | ret = btrfs_mksubvol(&file->f_path, name, namelen, |
6f72c7e2 | 1659 | NULL, transid, readonly, inherit); |
cb8e7090 | 1660 | } else { |
2903ff01 | 1661 | struct fd src = fdget(fd); |
3de4586c | 1662 | struct inode *src_inode; |
2903ff01 | 1663 | if (!src.file) { |
3de4586c | 1664 | ret = -EINVAL; |
a874a63e | 1665 | goto out_drop_write; |
3de4586c CM |
1666 | } |
1667 | ||
496ad9aa AV |
1668 | src_inode = file_inode(src.file); |
1669 | if (src_inode->i_sb != file_inode(file)->i_sb) { | |
c79b4713 | 1670 | btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, |
efe120a0 | 1671 | "Snapshot src from another FS"); |
23ad5b17 | 1672 | ret = -EXDEV; |
d0242061 DS |
1673 | } else if (!inode_owner_or_capable(src_inode)) { |
1674 | /* | |
1675 | * Subvolume creation is not restricted, but snapshots | |
1676 | * are limited to own subvolumes only | |
1677 | */ | |
1678 | ret = -EPERM; | |
ecd18815 AV |
1679 | } else { |
1680 | ret = btrfs_mksubvol(&file->f_path, name, namelen, | |
1681 | BTRFS_I(src_inode)->root, | |
1682 | transid, readonly, inherit); | |
3de4586c | 1683 | } |
2903ff01 | 1684 | fdput(src); |
cb8e7090 | 1685 | } |
a874a63e LB |
1686 | out_drop_write: |
1687 | mnt_drop_write_file(file); | |
f46b5a66 | 1688 | out: |
72fd032e SW |
1689 | return ret; |
1690 | } | |
1691 | ||
1692 | static noinline int btrfs_ioctl_snap_create(struct file *file, | |
fa0d2b9b | 1693 | void __user *arg, int subvol) |
72fd032e | 1694 | { |
fa0d2b9b | 1695 | struct btrfs_ioctl_vol_args *vol_args; |
72fd032e SW |
1696 | int ret; |
1697 | ||
325c50e3 JM |
1698 | if (!S_ISDIR(file_inode(file)->i_mode)) |
1699 | return -ENOTDIR; | |
1700 | ||
fa0d2b9b LZ |
1701 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1702 | if (IS_ERR(vol_args)) | |
1703 | return PTR_ERR(vol_args); | |
1704 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
72fd032e | 1705 | |
fa0d2b9b | 1706 | ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, |
b83cc969 | 1707 | vol_args->fd, subvol, |
6f72c7e2 | 1708 | NULL, false, NULL); |
fdfb1e4f | 1709 | |
fa0d2b9b LZ |
1710 | kfree(vol_args); |
1711 | return ret; | |
1712 | } | |
fdfb1e4f | 1713 | |
fa0d2b9b LZ |
1714 | static noinline int btrfs_ioctl_snap_create_v2(struct file *file, |
1715 | void __user *arg, int subvol) | |
1716 | { | |
1717 | struct btrfs_ioctl_vol_args_v2 *vol_args; | |
1718 | int ret; | |
1719 | u64 transid = 0; | |
1720 | u64 *ptr = NULL; | |
b83cc969 | 1721 | bool readonly = false; |
6f72c7e2 | 1722 | struct btrfs_qgroup_inherit *inherit = NULL; |
75eaa0e2 | 1723 | |
325c50e3 JM |
1724 | if (!S_ISDIR(file_inode(file)->i_mode)) |
1725 | return -ENOTDIR; | |
1726 | ||
fa0d2b9b LZ |
1727 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1728 | if (IS_ERR(vol_args)) | |
1729 | return PTR_ERR(vol_args); | |
1730 | vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; | |
75eaa0e2 | 1731 | |
b83cc969 | 1732 | if (vol_args->flags & |
6f72c7e2 AJ |
1733 | ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | |
1734 | BTRFS_SUBVOL_QGROUP_INHERIT)) { | |
b83cc969 | 1735 | ret = -EOPNOTSUPP; |
c47ca32d | 1736 | goto free_args; |
72fd032e | 1737 | } |
fa0d2b9b LZ |
1738 | |
1739 | if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) | |
1740 | ptr = &transid; | |
b83cc969 LZ |
1741 | if (vol_args->flags & BTRFS_SUBVOL_RDONLY) |
1742 | readonly = true; | |
6f72c7e2 | 1743 | if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { |
09cbfeaf | 1744 | if (vol_args->size > PAGE_SIZE) { |
6f72c7e2 | 1745 | ret = -EINVAL; |
c47ca32d | 1746 | goto free_args; |
6f72c7e2 AJ |
1747 | } |
1748 | inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); | |
1749 | if (IS_ERR(inherit)) { | |
1750 | ret = PTR_ERR(inherit); | |
c47ca32d | 1751 | goto free_args; |
6f72c7e2 AJ |
1752 | } |
1753 | } | |
fa0d2b9b LZ |
1754 | |
1755 | ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, | |
6f72c7e2 | 1756 | vol_args->fd, subvol, ptr, |
8696c533 | 1757 | readonly, inherit); |
c47ca32d DC |
1758 | if (ret) |
1759 | goto free_inherit; | |
fa0d2b9b | 1760 | |
c47ca32d DC |
1761 | if (ptr && copy_to_user(arg + |
1762 | offsetof(struct btrfs_ioctl_vol_args_v2, | |
1763 | transid), | |
1764 | ptr, sizeof(*ptr))) | |
fa0d2b9b | 1765 | ret = -EFAULT; |
c47ca32d DC |
1766 | |
1767 | free_inherit: | |
6f72c7e2 | 1768 | kfree(inherit); |
c47ca32d DC |
1769 | free_args: |
1770 | kfree(vol_args); | |
f46b5a66 CH |
1771 | return ret; |
1772 | } | |
1773 | ||
0caa102d LZ |
1774 | static noinline int btrfs_ioctl_subvol_getflags(struct file *file, |
1775 | void __user *arg) | |
1776 | { | |
496ad9aa | 1777 | struct inode *inode = file_inode(file); |
0b246afa | 1778 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
0caa102d LZ |
1779 | struct btrfs_root *root = BTRFS_I(inode)->root; |
1780 | int ret = 0; | |
1781 | u64 flags = 0; | |
1782 | ||
33345d01 | 1783 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) |
0caa102d LZ |
1784 | return -EINVAL; |
1785 | ||
0b246afa | 1786 | down_read(&fs_info->subvol_sem); |
0caa102d LZ |
1787 | if (btrfs_root_readonly(root)) |
1788 | flags |= BTRFS_SUBVOL_RDONLY; | |
0b246afa | 1789 | up_read(&fs_info->subvol_sem); |
0caa102d LZ |
1790 | |
1791 | if (copy_to_user(arg, &flags, sizeof(flags))) | |
1792 | ret = -EFAULT; | |
1793 | ||
1794 | return ret; | |
1795 | } | |
1796 | ||
1797 | static noinline int btrfs_ioctl_subvol_setflags(struct file *file, | |
1798 | void __user *arg) | |
1799 | { | |
496ad9aa | 1800 | struct inode *inode = file_inode(file); |
0b246afa | 1801 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
0caa102d LZ |
1802 | struct btrfs_root *root = BTRFS_I(inode)->root; |
1803 | struct btrfs_trans_handle *trans; | |
1804 | u64 root_flags; | |
1805 | u64 flags; | |
1806 | int ret = 0; | |
1807 | ||
bd60ea0f DS |
1808 | if (!inode_owner_or_capable(inode)) |
1809 | return -EPERM; | |
1810 | ||
b9ca0664 LB |
1811 | ret = mnt_want_write_file(file); |
1812 | if (ret) | |
1813 | goto out; | |
0caa102d | 1814 | |
b9ca0664 LB |
1815 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { |
1816 | ret = -EINVAL; | |
1817 | goto out_drop_write; | |
1818 | } | |
0caa102d | 1819 | |
b9ca0664 LB |
1820 | if (copy_from_user(&flags, arg, sizeof(flags))) { |
1821 | ret = -EFAULT; | |
1822 | goto out_drop_write; | |
1823 | } | |
0caa102d | 1824 | |
b9ca0664 LB |
1825 | if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { |
1826 | ret = -EINVAL; | |
1827 | goto out_drop_write; | |
1828 | } | |
0caa102d | 1829 | |
b9ca0664 LB |
1830 | if (flags & ~BTRFS_SUBVOL_RDONLY) { |
1831 | ret = -EOPNOTSUPP; | |
1832 | goto out_drop_write; | |
1833 | } | |
0caa102d | 1834 | |
0b246afa | 1835 | down_write(&fs_info->subvol_sem); |
0caa102d LZ |
1836 | |
1837 | /* nothing to do */ | |
1838 | if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) | |
b9ca0664 | 1839 | goto out_drop_sem; |
0caa102d LZ |
1840 | |
1841 | root_flags = btrfs_root_flags(&root->root_item); | |
2c686537 | 1842 | if (flags & BTRFS_SUBVOL_RDONLY) { |
0caa102d LZ |
1843 | btrfs_set_root_flags(&root->root_item, |
1844 | root_flags | BTRFS_ROOT_SUBVOL_RDONLY); | |
2c686537 DS |
1845 | } else { |
1846 | /* | |
1847 | * Block RO -> RW transition if this subvolume is involved in | |
1848 | * send | |
1849 | */ | |
1850 | spin_lock(&root->root_item_lock); | |
1851 | if (root->send_in_progress == 0) { | |
1852 | btrfs_set_root_flags(&root->root_item, | |
0caa102d | 1853 | root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); |
2c686537 DS |
1854 | spin_unlock(&root->root_item_lock); |
1855 | } else { | |
1856 | spin_unlock(&root->root_item_lock); | |
0b246afa JM |
1857 | btrfs_warn(fs_info, |
1858 | "Attempt to set subvolume %llu read-write during send", | |
1859 | root->root_key.objectid); | |
2c686537 DS |
1860 | ret = -EPERM; |
1861 | goto out_drop_sem; | |
1862 | } | |
1863 | } | |
0caa102d LZ |
1864 | |
1865 | trans = btrfs_start_transaction(root, 1); | |
1866 | if (IS_ERR(trans)) { | |
1867 | ret = PTR_ERR(trans); | |
1868 | goto out_reset; | |
1869 | } | |
1870 | ||
0b246afa | 1871 | ret = btrfs_update_root(trans, fs_info->tree_root, |
0caa102d LZ |
1872 | &root->root_key, &root->root_item); |
1873 | ||
3a45bb20 | 1874 | btrfs_commit_transaction(trans); |
0caa102d LZ |
1875 | out_reset: |
1876 | if (ret) | |
1877 | btrfs_set_root_flags(&root->root_item, root_flags); | |
b9ca0664 | 1878 | out_drop_sem: |
0b246afa | 1879 | up_write(&fs_info->subvol_sem); |
b9ca0664 LB |
1880 | out_drop_write: |
1881 | mnt_drop_write_file(file); | |
1882 | out: | |
0caa102d LZ |
1883 | return ret; |
1884 | } | |
1885 | ||
76dda93c YZ |
1886 | /* |
1887 | * helper to check if the subvolume references other subvolumes | |
1888 | */ | |
1889 | static noinline int may_destroy_subvol(struct btrfs_root *root) | |
1890 | { | |
0b246afa | 1891 | struct btrfs_fs_info *fs_info = root->fs_info; |
76dda93c | 1892 | struct btrfs_path *path; |
175a2b87 | 1893 | struct btrfs_dir_item *di; |
76dda93c | 1894 | struct btrfs_key key; |
175a2b87 | 1895 | u64 dir_id; |
76dda93c YZ |
1896 | int ret; |
1897 | ||
1898 | path = btrfs_alloc_path(); | |
1899 | if (!path) | |
1900 | return -ENOMEM; | |
1901 | ||
175a2b87 | 1902 | /* Make sure this root isn't set as the default subvol */ |
0b246afa JM |
1903 | dir_id = btrfs_super_root_dir(fs_info->super_copy); |
1904 | di = btrfs_lookup_dir_item(NULL, fs_info->tree_root, path, | |
175a2b87 JB |
1905 | dir_id, "default", 7, 0); |
1906 | if (di && !IS_ERR(di)) { | |
1907 | btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key); | |
1908 | if (key.objectid == root->root_key.objectid) { | |
72de6b53 | 1909 | ret = -EPERM; |
0b246afa | 1910 | btrfs_err(fs_info, |
5d163e0e JM |
1911 | "deleting default subvolume %llu is not allowed", |
1912 | key.objectid); | |
175a2b87 JB |
1913 | goto out; |
1914 | } | |
1915 | btrfs_release_path(path); | |
1916 | } | |
1917 | ||
76dda93c YZ |
1918 | key.objectid = root->root_key.objectid; |
1919 | key.type = BTRFS_ROOT_REF_KEY; | |
1920 | key.offset = (u64)-1; | |
1921 | ||
0b246afa | 1922 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); |
76dda93c YZ |
1923 | if (ret < 0) |
1924 | goto out; | |
1925 | BUG_ON(ret == 0); | |
1926 | ||
1927 | ret = 0; | |
1928 | if (path->slots[0] > 0) { | |
1929 | path->slots[0]--; | |
1930 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); | |
1931 | if (key.objectid == root->root_key.objectid && | |
1932 | key.type == BTRFS_ROOT_REF_KEY) | |
1933 | ret = -ENOTEMPTY; | |
1934 | } | |
1935 | out: | |
1936 | btrfs_free_path(path); | |
1937 | return ret; | |
1938 | } | |
1939 | ||
ac8e9819 CM |
1940 | static noinline int key_in_sk(struct btrfs_key *key, |
1941 | struct btrfs_ioctl_search_key *sk) | |
1942 | { | |
abc6e134 CM |
1943 | struct btrfs_key test; |
1944 | int ret; | |
1945 | ||
1946 | test.objectid = sk->min_objectid; | |
1947 | test.type = sk->min_type; | |
1948 | test.offset = sk->min_offset; | |
1949 | ||
1950 | ret = btrfs_comp_cpu_keys(key, &test); | |
1951 | if (ret < 0) | |
ac8e9819 | 1952 | return 0; |
abc6e134 CM |
1953 | |
1954 | test.objectid = sk->max_objectid; | |
1955 | test.type = sk->max_type; | |
1956 | test.offset = sk->max_offset; | |
1957 | ||
1958 | ret = btrfs_comp_cpu_keys(key, &test); | |
1959 | if (ret > 0) | |
ac8e9819 CM |
1960 | return 0; |
1961 | return 1; | |
1962 | } | |
1963 | ||
df397565 | 1964 | static noinline int copy_to_sk(struct btrfs_path *path, |
ac8e9819 CM |
1965 | struct btrfs_key *key, |
1966 | struct btrfs_ioctl_search_key *sk, | |
9b6e817d | 1967 | size_t *buf_size, |
ba346b35 | 1968 | char __user *ubuf, |
ac8e9819 CM |
1969 | unsigned long *sk_offset, |
1970 | int *num_found) | |
1971 | { | |
1972 | u64 found_transid; | |
1973 | struct extent_buffer *leaf; | |
1974 | struct btrfs_ioctl_search_header sh; | |
dd81d459 | 1975 | struct btrfs_key test; |
ac8e9819 CM |
1976 | unsigned long item_off; |
1977 | unsigned long item_len; | |
1978 | int nritems; | |
1979 | int i; | |
1980 | int slot; | |
ac8e9819 CM |
1981 | int ret = 0; |
1982 | ||
1983 | leaf = path->nodes[0]; | |
1984 | slot = path->slots[0]; | |
1985 | nritems = btrfs_header_nritems(leaf); | |
1986 | ||
1987 | if (btrfs_header_generation(leaf) > sk->max_transid) { | |
1988 | i = nritems; | |
1989 | goto advance_key; | |
1990 | } | |
1991 | found_transid = btrfs_header_generation(leaf); | |
1992 | ||
1993 | for (i = slot; i < nritems; i++) { | |
1994 | item_off = btrfs_item_ptr_offset(leaf, i); | |
1995 | item_len = btrfs_item_size_nr(leaf, i); | |
1996 | ||
03b71c6c GP |
1997 | btrfs_item_key_to_cpu(leaf, key, i); |
1998 | if (!key_in_sk(key, sk)) | |
1999 | continue; | |
2000 | ||
9b6e817d | 2001 | if (sizeof(sh) + item_len > *buf_size) { |
8f5f6178 GH |
2002 | if (*num_found) { |
2003 | ret = 1; | |
2004 | goto out; | |
2005 | } | |
2006 | ||
2007 | /* | |
2008 | * return one empty item back for v1, which does not | |
2009 | * handle -EOVERFLOW | |
2010 | */ | |
2011 | ||
9b6e817d | 2012 | *buf_size = sizeof(sh) + item_len; |
ac8e9819 | 2013 | item_len = 0; |
8f5f6178 GH |
2014 | ret = -EOVERFLOW; |
2015 | } | |
ac8e9819 | 2016 | |
9b6e817d | 2017 | if (sizeof(sh) + item_len + *sk_offset > *buf_size) { |
ac8e9819 | 2018 | ret = 1; |
25c9bc2e | 2019 | goto out; |
ac8e9819 CM |
2020 | } |
2021 | ||
ac8e9819 CM |
2022 | sh.objectid = key->objectid; |
2023 | sh.offset = key->offset; | |
2024 | sh.type = key->type; | |
2025 | sh.len = item_len; | |
2026 | sh.transid = found_transid; | |
2027 | ||
2028 | /* copy search result header */ | |
ba346b35 GH |
2029 | if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { |
2030 | ret = -EFAULT; | |
2031 | goto out; | |
2032 | } | |
2033 | ||
ac8e9819 CM |
2034 | *sk_offset += sizeof(sh); |
2035 | ||
2036 | if (item_len) { | |
ba346b35 | 2037 | char __user *up = ubuf + *sk_offset; |
ac8e9819 | 2038 | /* copy the item */ |
ba346b35 GH |
2039 | if (read_extent_buffer_to_user(leaf, up, |
2040 | item_off, item_len)) { | |
2041 | ret = -EFAULT; | |
2042 | goto out; | |
2043 | } | |
2044 | ||
ac8e9819 | 2045 | *sk_offset += item_len; |
ac8e9819 | 2046 | } |
e2156867 | 2047 | (*num_found)++; |
ac8e9819 | 2048 | |
8f5f6178 GH |
2049 | if (ret) /* -EOVERFLOW from above */ |
2050 | goto out; | |
2051 | ||
25c9bc2e GH |
2052 | if (*num_found >= sk->nr_items) { |
2053 | ret = 1; | |
2054 | goto out; | |
2055 | } | |
ac8e9819 CM |
2056 | } |
2057 | advance_key: | |
abc6e134 | 2058 | ret = 0; |
dd81d459 NA |
2059 | test.objectid = sk->max_objectid; |
2060 | test.type = sk->max_type; | |
2061 | test.offset = sk->max_offset; | |
2062 | if (btrfs_comp_cpu_keys(key, &test) >= 0) | |
2063 | ret = 1; | |
2064 | else if (key->offset < (u64)-1) | |
ac8e9819 | 2065 | key->offset++; |
dd81d459 | 2066 | else if (key->type < (u8)-1) { |
abc6e134 | 2067 | key->offset = 0; |
ac8e9819 | 2068 | key->type++; |
dd81d459 | 2069 | } else if (key->objectid < (u64)-1) { |
abc6e134 CM |
2070 | key->offset = 0; |
2071 | key->type = 0; | |
ac8e9819 | 2072 | key->objectid++; |
abc6e134 CM |
2073 | } else |
2074 | ret = 1; | |
25c9bc2e | 2075 | out: |
ba346b35 GH |
2076 | /* |
2077 | * 0: all items from this leaf copied, continue with next | |
2078 | * 1: * more items can be copied, but unused buffer is too small | |
2079 | * * all items were found | |
2080 | * Either way, it will stops the loop which iterates to the next | |
2081 | * leaf | |
2082 | * -EOVERFLOW: item was to large for buffer | |
2083 | * -EFAULT: could not copy extent buffer back to userspace | |
2084 | */ | |
ac8e9819 CM |
2085 | return ret; |
2086 | } | |
2087 | ||
2088 | static noinline int search_ioctl(struct inode *inode, | |
12544442 | 2089 | struct btrfs_ioctl_search_key *sk, |
9b6e817d | 2090 | size_t *buf_size, |
ba346b35 | 2091 | char __user *ubuf) |
ac8e9819 | 2092 | { |
0b246afa | 2093 | struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); |
ac8e9819 CM |
2094 | struct btrfs_root *root; |
2095 | struct btrfs_key key; | |
ac8e9819 | 2096 | struct btrfs_path *path; |
ac8e9819 CM |
2097 | int ret; |
2098 | int num_found = 0; | |
2099 | unsigned long sk_offset = 0; | |
2100 | ||
9b6e817d GH |
2101 | if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { |
2102 | *buf_size = sizeof(struct btrfs_ioctl_search_header); | |
12544442 | 2103 | return -EOVERFLOW; |
9b6e817d | 2104 | } |
12544442 | 2105 | |
ac8e9819 CM |
2106 | path = btrfs_alloc_path(); |
2107 | if (!path) | |
2108 | return -ENOMEM; | |
2109 | ||
2110 | if (sk->tree_id == 0) { | |
2111 | /* search the root of the inode that was passed */ | |
2112 | root = BTRFS_I(inode)->root; | |
2113 | } else { | |
2114 | key.objectid = sk->tree_id; | |
2115 | key.type = BTRFS_ROOT_ITEM_KEY; | |
2116 | key.offset = (u64)-1; | |
2117 | root = btrfs_read_fs_root_no_name(info, &key); | |
2118 | if (IS_ERR(root)) { | |
ac8e9819 CM |
2119 | btrfs_free_path(path); |
2120 | return -ENOENT; | |
2121 | } | |
2122 | } | |
2123 | ||
2124 | key.objectid = sk->min_objectid; | |
2125 | key.type = sk->min_type; | |
2126 | key.offset = sk->min_offset; | |
2127 | ||
67871254 | 2128 | while (1) { |
6174d3cb | 2129 | ret = btrfs_search_forward(root, &key, path, sk->min_transid); |
ac8e9819 CM |
2130 | if (ret != 0) { |
2131 | if (ret > 0) | |
2132 | ret = 0; | |
2133 | goto err; | |
2134 | } | |
df397565 | 2135 | ret = copy_to_sk(path, &key, sk, buf_size, ubuf, |
ac8e9819 | 2136 | &sk_offset, &num_found); |
b3b4aa74 | 2137 | btrfs_release_path(path); |
25c9bc2e | 2138 | if (ret) |
ac8e9819 CM |
2139 | break; |
2140 | ||
2141 | } | |
8f5f6178 GH |
2142 | if (ret > 0) |
2143 | ret = 0; | |
ac8e9819 CM |
2144 | err: |
2145 | sk->nr_items = num_found; | |
2146 | btrfs_free_path(path); | |
2147 | return ret; | |
2148 | } | |
2149 | ||
2150 | static noinline int btrfs_ioctl_tree_search(struct file *file, | |
2151 | void __user *argp) | |
2152 | { | |
ba346b35 GH |
2153 | struct btrfs_ioctl_search_args __user *uargs; |
2154 | struct btrfs_ioctl_search_key sk; | |
9b6e817d GH |
2155 | struct inode *inode; |
2156 | int ret; | |
2157 | size_t buf_size; | |
ac8e9819 CM |
2158 | |
2159 | if (!capable(CAP_SYS_ADMIN)) | |
2160 | return -EPERM; | |
2161 | ||
ba346b35 GH |
2162 | uargs = (struct btrfs_ioctl_search_args __user *)argp; |
2163 | ||
2164 | if (copy_from_user(&sk, &uargs->key, sizeof(sk))) | |
2165 | return -EFAULT; | |
ac8e9819 | 2166 | |
ba346b35 | 2167 | buf_size = sizeof(uargs->buf); |
ac8e9819 | 2168 | |
496ad9aa | 2169 | inode = file_inode(file); |
ba346b35 | 2170 | ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); |
8f5f6178 GH |
2171 | |
2172 | /* | |
2173 | * In the origin implementation an overflow is handled by returning a | |
2174 | * search header with a len of zero, so reset ret. | |
2175 | */ | |
2176 | if (ret == -EOVERFLOW) | |
2177 | ret = 0; | |
2178 | ||
ba346b35 | 2179 | if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) |
ac8e9819 | 2180 | ret = -EFAULT; |
ac8e9819 CM |
2181 | return ret; |
2182 | } | |
2183 | ||
cc68a8a5 GH |
2184 | static noinline int btrfs_ioctl_tree_search_v2(struct file *file, |
2185 | void __user *argp) | |
2186 | { | |
2187 | struct btrfs_ioctl_search_args_v2 __user *uarg; | |
2188 | struct btrfs_ioctl_search_args_v2 args; | |
2189 | struct inode *inode; | |
2190 | int ret; | |
2191 | size_t buf_size; | |
ee22184b | 2192 | const size_t buf_limit = SZ_16M; |
cc68a8a5 GH |
2193 | |
2194 | if (!capable(CAP_SYS_ADMIN)) | |
2195 | return -EPERM; | |
2196 | ||
2197 | /* copy search header and buffer size */ | |
2198 | uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; | |
2199 | if (copy_from_user(&args, uarg, sizeof(args))) | |
2200 | return -EFAULT; | |
2201 | ||
2202 | buf_size = args.buf_size; | |
2203 | ||
2204 | if (buf_size < sizeof(struct btrfs_ioctl_search_header)) | |
2205 | return -EOVERFLOW; | |
2206 | ||
2207 | /* limit result size to 16MB */ | |
2208 | if (buf_size > buf_limit) | |
2209 | buf_size = buf_limit; | |
2210 | ||
2211 | inode = file_inode(file); | |
2212 | ret = search_ioctl(inode, &args.key, &buf_size, | |
2213 | (char *)(&uarg->buf[0])); | |
2214 | if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) | |
2215 | ret = -EFAULT; | |
2216 | else if (ret == -EOVERFLOW && | |
2217 | copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) | |
2218 | ret = -EFAULT; | |
2219 | ||
ac8e9819 CM |
2220 | return ret; |
2221 | } | |
2222 | ||
98d377a0 | 2223 | /* |
ac8e9819 CM |
2224 | * Search INODE_REFs to identify path name of 'dirid' directory |
2225 | * in a 'tree_id' tree. and sets path name to 'name'. | |
2226 | */ | |
98d377a0 TH |
2227 | static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, |
2228 | u64 tree_id, u64 dirid, char *name) | |
2229 | { | |
2230 | struct btrfs_root *root; | |
2231 | struct btrfs_key key; | |
ac8e9819 | 2232 | char *ptr; |
98d377a0 TH |
2233 | int ret = -1; |
2234 | int slot; | |
2235 | int len; | |
2236 | int total_len = 0; | |
2237 | struct btrfs_inode_ref *iref; | |
2238 | struct extent_buffer *l; | |
2239 | struct btrfs_path *path; | |
2240 | ||
2241 | if (dirid == BTRFS_FIRST_FREE_OBJECTID) { | |
2242 | name[0]='\0'; | |
2243 | return 0; | |
2244 | } | |
2245 | ||
2246 | path = btrfs_alloc_path(); | |
2247 | if (!path) | |
2248 | return -ENOMEM; | |
2249 | ||
ac8e9819 | 2250 | ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; |
98d377a0 TH |
2251 | |
2252 | key.objectid = tree_id; | |
2253 | key.type = BTRFS_ROOT_ITEM_KEY; | |
2254 | key.offset = (u64)-1; | |
2255 | root = btrfs_read_fs_root_no_name(info, &key); | |
2256 | if (IS_ERR(root)) { | |
f14d104d | 2257 | btrfs_err(info, "could not find root %llu", tree_id); |
8ad6fcab CM |
2258 | ret = -ENOENT; |
2259 | goto out; | |
98d377a0 TH |
2260 | } |
2261 | ||
2262 | key.objectid = dirid; | |
2263 | key.type = BTRFS_INODE_REF_KEY; | |
8ad6fcab | 2264 | key.offset = (u64)-1; |
98d377a0 | 2265 | |
67871254 | 2266 | while (1) { |
98d377a0 TH |
2267 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
2268 | if (ret < 0) | |
2269 | goto out; | |
18674c6c FDBM |
2270 | else if (ret > 0) { |
2271 | ret = btrfs_previous_item(root, path, dirid, | |
2272 | BTRFS_INODE_REF_KEY); | |
2273 | if (ret < 0) | |
2274 | goto out; | |
2275 | else if (ret > 0) { | |
2276 | ret = -ENOENT; | |
2277 | goto out; | |
2278 | } | |
2279 | } | |
98d377a0 TH |
2280 | |
2281 | l = path->nodes[0]; | |
2282 | slot = path->slots[0]; | |
2283 | btrfs_item_key_to_cpu(l, &key, slot); | |
2284 | ||
98d377a0 TH |
2285 | iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); |
2286 | len = btrfs_inode_ref_name_len(l, iref); | |
2287 | ptr -= len + 1; | |
2288 | total_len += len + 1; | |
a696cf35 FDBM |
2289 | if (ptr < name) { |
2290 | ret = -ENAMETOOLONG; | |
98d377a0 | 2291 | goto out; |
a696cf35 | 2292 | } |
98d377a0 TH |
2293 | |
2294 | *(ptr + len) = '/'; | |
67871254 | 2295 | read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); |
98d377a0 TH |
2296 | |
2297 | if (key.offset == BTRFS_FIRST_FREE_OBJECTID) | |
2298 | break; | |
2299 | ||
b3b4aa74 | 2300 | btrfs_release_path(path); |
98d377a0 | 2301 | key.objectid = key.offset; |
8ad6fcab | 2302 | key.offset = (u64)-1; |
98d377a0 | 2303 | dirid = key.objectid; |
98d377a0 | 2304 | } |
77906a50 | 2305 | memmove(name, ptr, total_len); |
67871254 | 2306 | name[total_len] = '\0'; |
98d377a0 TH |
2307 | ret = 0; |
2308 | out: | |
2309 | btrfs_free_path(path); | |
ac8e9819 CM |
2310 | return ret; |
2311 | } | |
2312 | ||
2313 | static noinline int btrfs_ioctl_ino_lookup(struct file *file, | |
2314 | void __user *argp) | |
2315 | { | |
2316 | struct btrfs_ioctl_ino_lookup_args *args; | |
2317 | struct inode *inode; | |
01b810b8 | 2318 | int ret = 0; |
ac8e9819 | 2319 | |
2354d08f JL |
2320 | args = memdup_user(argp, sizeof(*args)); |
2321 | if (IS_ERR(args)) | |
2322 | return PTR_ERR(args); | |
c2b96929 | 2323 | |
496ad9aa | 2324 | inode = file_inode(file); |
ac8e9819 | 2325 | |
01b810b8 DS |
2326 | /* |
2327 | * Unprivileged query to obtain the containing subvolume root id. The | |
2328 | * path is reset so it's consistent with btrfs_search_path_in_tree. | |
2329 | */ | |
1b53ac4d CM |
2330 | if (args->treeid == 0) |
2331 | args->treeid = BTRFS_I(inode)->root->root_key.objectid; | |
2332 | ||
01b810b8 DS |
2333 | if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { |
2334 | args->name[0] = 0; | |
2335 | goto out; | |
2336 | } | |
2337 | ||
2338 | if (!capable(CAP_SYS_ADMIN)) { | |
2339 | ret = -EPERM; | |
2340 | goto out; | |
2341 | } | |
2342 | ||
ac8e9819 CM |
2343 | ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, |
2344 | args->treeid, args->objectid, | |
2345 | args->name); | |
2346 | ||
01b810b8 | 2347 | out: |
ac8e9819 CM |
2348 | if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) |
2349 | ret = -EFAULT; | |
2350 | ||
2351 | kfree(args); | |
98d377a0 TH |
2352 | return ret; |
2353 | } | |
2354 | ||
76dda93c YZ |
2355 | static noinline int btrfs_ioctl_snap_destroy(struct file *file, |
2356 | void __user *arg) | |
2357 | { | |
54563d41 | 2358 | struct dentry *parent = file->f_path.dentry; |
0b246afa | 2359 | struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); |
76dda93c | 2360 | struct dentry *dentry; |
2b0143b5 | 2361 | struct inode *dir = d_inode(parent); |
76dda93c YZ |
2362 | struct inode *inode; |
2363 | struct btrfs_root *root = BTRFS_I(dir)->root; | |
2364 | struct btrfs_root *dest = NULL; | |
2365 | struct btrfs_ioctl_vol_args *vol_args; | |
2366 | struct btrfs_trans_handle *trans; | |
c58aaad2 | 2367 | struct btrfs_block_rsv block_rsv; |
521e0546 | 2368 | u64 root_flags; |
c58aaad2 | 2369 | u64 qgroup_reserved; |
76dda93c YZ |
2370 | int namelen; |
2371 | int ret; | |
2372 | int err = 0; | |
2373 | ||
325c50e3 JM |
2374 | if (!S_ISDIR(dir->i_mode)) |
2375 | return -ENOTDIR; | |
2376 | ||
76dda93c YZ |
2377 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
2378 | if (IS_ERR(vol_args)) | |
2379 | return PTR_ERR(vol_args); | |
2380 | ||
2381 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
2382 | namelen = strlen(vol_args->name); | |
2383 | if (strchr(vol_args->name, '/') || | |
2384 | strncmp(vol_args->name, "..", namelen) == 0) { | |
2385 | err = -EINVAL; | |
2386 | goto out; | |
2387 | } | |
2388 | ||
a561be71 | 2389 | err = mnt_want_write_file(file); |
76dda93c YZ |
2390 | if (err) |
2391 | goto out; | |
2392 | ||
521e0546 | 2393 | |
00235411 AV |
2394 | err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); |
2395 | if (err == -EINTR) | |
2396 | goto out_drop_write; | |
76dda93c YZ |
2397 | dentry = lookup_one_len(vol_args->name, parent, namelen); |
2398 | if (IS_ERR(dentry)) { | |
2399 | err = PTR_ERR(dentry); | |
2400 | goto out_unlock_dir; | |
2401 | } | |
2402 | ||
2b0143b5 | 2403 | if (d_really_is_negative(dentry)) { |
76dda93c YZ |
2404 | err = -ENOENT; |
2405 | goto out_dput; | |
2406 | } | |
2407 | ||
2b0143b5 | 2408 | inode = d_inode(dentry); |
4260f7c7 | 2409 | dest = BTRFS_I(inode)->root; |
67871254 | 2410 | if (!capable(CAP_SYS_ADMIN)) { |
4260f7c7 SW |
2411 | /* |
2412 | * Regular user. Only allow this with a special mount | |
2413 | * option, when the user has write+exec access to the | |
2414 | * subvol root, and when rmdir(2) would have been | |
2415 | * allowed. | |
2416 | * | |
2417 | * Note that this is _not_ check that the subvol is | |
2418 | * empty or doesn't contain data that we wouldn't | |
2419 | * otherwise be able to delete. | |
2420 | * | |
2421 | * Users who want to delete empty subvols should try | |
2422 | * rmdir(2). | |
2423 | */ | |
2424 | err = -EPERM; | |
0b246afa | 2425 | if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) |
4260f7c7 SW |
2426 | goto out_dput; |
2427 | ||
2428 | /* | |
2429 | * Do not allow deletion if the parent dir is the same | |
2430 | * as the dir to be deleted. That means the ioctl | |
2431 | * must be called on the dentry referencing the root | |
2432 | * of the subvol, not a random directory contained | |
2433 | * within it. | |
2434 | */ | |
2435 | err = -EINVAL; | |
2436 | if (root == dest) | |
2437 | goto out_dput; | |
2438 | ||
2439 | err = inode_permission(inode, MAY_WRITE | MAY_EXEC); | |
2440 | if (err) | |
2441 | goto out_dput; | |
4260f7c7 SW |
2442 | } |
2443 | ||
5c39da5b MX |
2444 | /* check if subvolume may be deleted by a user */ |
2445 | err = btrfs_may_delete(dir, dentry, 1); | |
2446 | if (err) | |
2447 | goto out_dput; | |
2448 | ||
33345d01 | 2449 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { |
76dda93c YZ |
2450 | err = -EINVAL; |
2451 | goto out_dput; | |
2452 | } | |
2453 | ||
5955102c | 2454 | inode_lock(inode); |
521e0546 DS |
2455 | |
2456 | /* | |
2457 | * Don't allow to delete a subvolume with send in progress. This is | |
2458 | * inside the i_mutex so the error handling that has to drop the bit | |
2459 | * again is not run concurrently. | |
2460 | */ | |
2461 | spin_lock(&dest->root_item_lock); | |
c55bfa67 FM |
2462 | root_flags = btrfs_root_flags(&dest->root_item); |
2463 | if (dest->send_in_progress == 0) { | |
2464 | btrfs_set_root_flags(&dest->root_item, | |
521e0546 DS |
2465 | root_flags | BTRFS_ROOT_SUBVOL_DEAD); |
2466 | spin_unlock(&dest->root_item_lock); | |
2467 | } else { | |
2468 | spin_unlock(&dest->root_item_lock); | |
0b246afa JM |
2469 | btrfs_warn(fs_info, |
2470 | "Attempt to delete subvolume %llu during send", | |
2471 | dest->root_key.objectid); | |
521e0546 | 2472 | err = -EPERM; |
909e26dc | 2473 | goto out_unlock_inode; |
521e0546 DS |
2474 | } |
2475 | ||
0b246afa | 2476 | down_write(&fs_info->subvol_sem); |
76dda93c YZ |
2477 | |
2478 | err = may_destroy_subvol(dest); | |
2479 | if (err) | |
2480 | goto out_up_write; | |
2481 | ||
c58aaad2 MX |
2482 | btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); |
2483 | /* | |
2484 | * One for dir inode, two for dir entries, two for root | |
2485 | * ref/backref. | |
2486 | */ | |
2487 | err = btrfs_subvolume_reserve_metadata(root, &block_rsv, | |
ee3441b4 | 2488 | 5, &qgroup_reserved, true); |
c58aaad2 MX |
2489 | if (err) |
2490 | goto out_up_write; | |
2491 | ||
a22285a6 YZ |
2492 | trans = btrfs_start_transaction(root, 0); |
2493 | if (IS_ERR(trans)) { | |
2494 | err = PTR_ERR(trans); | |
c58aaad2 | 2495 | goto out_release; |
a22285a6 | 2496 | } |
c58aaad2 MX |
2497 | trans->block_rsv = &block_rsv; |
2498 | trans->bytes_reserved = block_rsv.size; | |
a22285a6 | 2499 | |
2be63d5c FM |
2500 | btrfs_record_snapshot_destroy(trans, dir); |
2501 | ||
76dda93c YZ |
2502 | ret = btrfs_unlink_subvol(trans, root, dir, |
2503 | dest->root_key.objectid, | |
2504 | dentry->d_name.name, | |
2505 | dentry->d_name.len); | |
79787eaa JM |
2506 | if (ret) { |
2507 | err = ret; | |
66642832 | 2508 | btrfs_abort_transaction(trans, ret); |
79787eaa JM |
2509 | goto out_end_trans; |
2510 | } | |
76dda93c YZ |
2511 | |
2512 | btrfs_record_root_in_trans(trans, dest); | |
2513 | ||
2514 | memset(&dest->root_item.drop_progress, 0, | |
2515 | sizeof(dest->root_item.drop_progress)); | |
2516 | dest->root_item.drop_level = 0; | |
2517 | btrfs_set_root_refs(&dest->root_item, 0); | |
2518 | ||
27cdeb70 | 2519 | if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) { |
d68fc57b | 2520 | ret = btrfs_insert_orphan_item(trans, |
0b246afa | 2521 | fs_info->tree_root, |
d68fc57b | 2522 | dest->root_key.objectid); |
79787eaa | 2523 | if (ret) { |
66642832 | 2524 | btrfs_abort_transaction(trans, ret); |
79787eaa JM |
2525 | err = ret; |
2526 | goto out_end_trans; | |
2527 | } | |
d68fc57b | 2528 | } |
dd5f9615 | 2529 | |
0b246afa | 2530 | ret = btrfs_uuid_tree_rem(trans, fs_info, dest->root_item.uuid, |
6bccf3ab | 2531 | BTRFS_UUID_KEY_SUBVOL, |
dd5f9615 SB |
2532 | dest->root_key.objectid); |
2533 | if (ret && ret != -ENOENT) { | |
66642832 | 2534 | btrfs_abort_transaction(trans, ret); |
dd5f9615 SB |
2535 | err = ret; |
2536 | goto out_end_trans; | |
2537 | } | |
2538 | if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) { | |
0b246afa | 2539 | ret = btrfs_uuid_tree_rem(trans, fs_info, |
dd5f9615 SB |
2540 | dest->root_item.received_uuid, |
2541 | BTRFS_UUID_KEY_RECEIVED_SUBVOL, | |
2542 | dest->root_key.objectid); | |
2543 | if (ret && ret != -ENOENT) { | |
66642832 | 2544 | btrfs_abort_transaction(trans, ret); |
dd5f9615 SB |
2545 | err = ret; |
2546 | goto out_end_trans; | |
2547 | } | |
2548 | } | |
2549 | ||
79787eaa | 2550 | out_end_trans: |
c58aaad2 MX |
2551 | trans->block_rsv = NULL; |
2552 | trans->bytes_reserved = 0; | |
3a45bb20 | 2553 | ret = btrfs_end_transaction(trans); |
79787eaa JM |
2554 | if (ret && !err) |
2555 | err = ret; | |
76dda93c | 2556 | inode->i_flags |= S_DEAD; |
c58aaad2 | 2557 | out_release: |
2ff7e61e | 2558 | btrfs_subvolume_release_metadata(fs_info, &block_rsv, qgroup_reserved); |
76dda93c | 2559 | out_up_write: |
0b246afa | 2560 | up_write(&fs_info->subvol_sem); |
521e0546 DS |
2561 | if (err) { |
2562 | spin_lock(&dest->root_item_lock); | |
c55bfa67 FM |
2563 | root_flags = btrfs_root_flags(&dest->root_item); |
2564 | btrfs_set_root_flags(&dest->root_item, | |
521e0546 DS |
2565 | root_flags & ~BTRFS_ROOT_SUBVOL_DEAD); |
2566 | spin_unlock(&dest->root_item_lock); | |
2567 | } | |
909e26dc | 2568 | out_unlock_inode: |
5955102c | 2569 | inode_unlock(inode); |
76dda93c | 2570 | if (!err) { |
64ad6c48 | 2571 | d_invalidate(dentry); |
76dda93c YZ |
2572 | btrfs_invalidate_inodes(dest); |
2573 | d_delete(dentry); | |
61155aa0 | 2574 | ASSERT(dest->send_in_progress == 0); |
fa6ac876 LB |
2575 | |
2576 | /* the last ref */ | |
57cdc8db DS |
2577 | if (dest->ino_cache_inode) { |
2578 | iput(dest->ino_cache_inode); | |
2579 | dest->ino_cache_inode = NULL; | |
fa6ac876 | 2580 | } |
76dda93c YZ |
2581 | } |
2582 | out_dput: | |
2583 | dput(dentry); | |
2584 | out_unlock_dir: | |
5955102c | 2585 | inode_unlock(dir); |
00235411 | 2586 | out_drop_write: |
2a79f17e | 2587 | mnt_drop_write_file(file); |
76dda93c YZ |
2588 | out: |
2589 | kfree(vol_args); | |
2590 | return err; | |
2591 | } | |
2592 | ||
1e701a32 | 2593 | static int btrfs_ioctl_defrag(struct file *file, void __user *argp) |
f46b5a66 | 2594 | { |
496ad9aa | 2595 | struct inode *inode = file_inode(file); |
f46b5a66 | 2596 | struct btrfs_root *root = BTRFS_I(inode)->root; |
1e701a32 | 2597 | struct btrfs_ioctl_defrag_range_args *range; |
c146afad YZ |
2598 | int ret; |
2599 | ||
25122d15 ID |
2600 | ret = mnt_want_write_file(file); |
2601 | if (ret) | |
2602 | return ret; | |
b83cc969 | 2603 | |
25122d15 ID |
2604 | if (btrfs_root_readonly(root)) { |
2605 | ret = -EROFS; | |
2606 | goto out; | |
5ac00add | 2607 | } |
f46b5a66 CH |
2608 | |
2609 | switch (inode->i_mode & S_IFMT) { | |
2610 | case S_IFDIR: | |
e441d54d CM |
2611 | if (!capable(CAP_SYS_ADMIN)) { |
2612 | ret = -EPERM; | |
2613 | goto out; | |
2614 | } | |
de78b51a | 2615 | ret = btrfs_defrag_root(root); |
f46b5a66 CH |
2616 | break; |
2617 | case S_IFREG: | |
e441d54d CM |
2618 | if (!(file->f_mode & FMODE_WRITE)) { |
2619 | ret = -EINVAL; | |
2620 | goto out; | |
2621 | } | |
1e701a32 CM |
2622 | |
2623 | range = kzalloc(sizeof(*range), GFP_KERNEL); | |
2624 | if (!range) { | |
2625 | ret = -ENOMEM; | |
2626 | goto out; | |
2627 | } | |
2628 | ||
2629 | if (argp) { | |
2630 | if (copy_from_user(range, argp, | |
2631 | sizeof(*range))) { | |
2632 | ret = -EFAULT; | |
2633 | kfree(range); | |
683be16e | 2634 | goto out; |
1e701a32 CM |
2635 | } |
2636 | /* compression requires us to start the IO */ | |
2637 | if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { | |
2638 | range->flags |= BTRFS_DEFRAG_RANGE_START_IO; | |
2639 | range->extent_thresh = (u32)-1; | |
2640 | } | |
2641 | } else { | |
2642 | /* the rest are all set to zero by kzalloc */ | |
2643 | range->len = (u64)-1; | |
2644 | } | |
496ad9aa | 2645 | ret = btrfs_defrag_file(file_inode(file), file, |
4cb5300b CM |
2646 | range, 0, 0); |
2647 | if (ret > 0) | |
2648 | ret = 0; | |
1e701a32 | 2649 | kfree(range); |
f46b5a66 | 2650 | break; |
8929ecfa YZ |
2651 | default: |
2652 | ret = -EINVAL; | |
f46b5a66 | 2653 | } |
e441d54d | 2654 | out: |
25122d15 | 2655 | mnt_drop_write_file(file); |
e441d54d | 2656 | return ret; |
f46b5a66 CH |
2657 | } |
2658 | ||
2ff7e61e | 2659 | static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) |
f46b5a66 CH |
2660 | { |
2661 | struct btrfs_ioctl_vol_args *vol_args; | |
2662 | int ret; | |
2663 | ||
e441d54d CM |
2664 | if (!capable(CAP_SYS_ADMIN)) |
2665 | return -EPERM; | |
2666 | ||
0b246afa | 2667 | if (atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) |
e57138b3 | 2668 | return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
c9e9f97b | 2669 | |
0b246afa | 2670 | mutex_lock(&fs_info->volume_mutex); |
dae7b665 | 2671 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
2672 | if (IS_ERR(vol_args)) { |
2673 | ret = PTR_ERR(vol_args); | |
2674 | goto out; | |
2675 | } | |
f46b5a66 | 2676 | |
5516e595 | 2677 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
2ff7e61e | 2678 | ret = btrfs_init_new_device(fs_info, vol_args->name); |
f46b5a66 | 2679 | |
43d20761 | 2680 | if (!ret) |
0b246afa | 2681 | btrfs_info(fs_info, "disk added %s", vol_args->name); |
43d20761 | 2682 | |
f46b5a66 | 2683 | kfree(vol_args); |
c9e9f97b | 2684 | out: |
0b246afa JM |
2685 | mutex_unlock(&fs_info->volume_mutex); |
2686 | atomic_set(&fs_info->mutually_exclusive_operation_running, 0); | |
f46b5a66 CH |
2687 | return ret; |
2688 | } | |
2689 | ||
6b526ed7 | 2690 | static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) |
f46b5a66 | 2691 | { |
0b246afa JM |
2692 | struct inode *inode = file_inode(file); |
2693 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
6b526ed7 | 2694 | struct btrfs_ioctl_vol_args_v2 *vol_args; |
f46b5a66 CH |
2695 | int ret; |
2696 | ||
e441d54d CM |
2697 | if (!capable(CAP_SYS_ADMIN)) |
2698 | return -EPERM; | |
2699 | ||
da24927b MX |
2700 | ret = mnt_want_write_file(file); |
2701 | if (ret) | |
2702 | return ret; | |
c146afad | 2703 | |
dae7b665 | 2704 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
2705 | if (IS_ERR(vol_args)) { |
2706 | ret = PTR_ERR(vol_args); | |
c47ca32d | 2707 | goto err_drop; |
c9e9f97b | 2708 | } |
f46b5a66 | 2709 | |
6b526ed7 | 2710 | /* Check for compatibility reject unknown flags */ |
735654ea DS |
2711 | if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) |
2712 | return -EOPNOTSUPP; | |
f46b5a66 | 2713 | |
0b246afa | 2714 | if (atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { |
183860f6 AJ |
2715 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
2716 | goto out; | |
2717 | } | |
2718 | ||
0b246afa | 2719 | mutex_lock(&fs_info->volume_mutex); |
735654ea | 2720 | if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { |
2ff7e61e | 2721 | ret = btrfs_rm_device(fs_info, NULL, vol_args->devid); |
6b526ed7 AJ |
2722 | } else { |
2723 | vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; | |
2ff7e61e | 2724 | ret = btrfs_rm_device(fs_info, vol_args->name, 0); |
6b526ed7 | 2725 | } |
0b246afa JM |
2726 | mutex_unlock(&fs_info->volume_mutex); |
2727 | atomic_set(&fs_info->mutually_exclusive_operation_running, 0); | |
183860f6 | 2728 | |
6b526ed7 | 2729 | if (!ret) { |
735654ea | 2730 | if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) |
0b246afa | 2731 | btrfs_info(fs_info, "device deleted: id %llu", |
6b526ed7 AJ |
2732 | vol_args->devid); |
2733 | else | |
0b246afa | 2734 | btrfs_info(fs_info, "device deleted: %s", |
6b526ed7 AJ |
2735 | vol_args->name); |
2736 | } | |
183860f6 AJ |
2737 | out: |
2738 | kfree(vol_args); | |
c47ca32d | 2739 | err_drop: |
4ac20c70 | 2740 | mnt_drop_write_file(file); |
f46b5a66 CH |
2741 | return ret; |
2742 | } | |
2743 | ||
da24927b | 2744 | static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) |
f46b5a66 | 2745 | { |
0b246afa JM |
2746 | struct inode *inode = file_inode(file); |
2747 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
f46b5a66 CH |
2748 | struct btrfs_ioctl_vol_args *vol_args; |
2749 | int ret; | |
2750 | ||
e441d54d CM |
2751 | if (!capable(CAP_SYS_ADMIN)) |
2752 | return -EPERM; | |
2753 | ||
da24927b MX |
2754 | ret = mnt_want_write_file(file); |
2755 | if (ret) | |
2756 | return ret; | |
c146afad | 2757 | |
0b246afa | 2758 | if (atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { |
183860f6 | 2759 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
58d7bbf8 DS |
2760 | goto out_drop_write; |
2761 | } | |
2762 | ||
2763 | vol_args = memdup_user(arg, sizeof(*vol_args)); | |
2764 | if (IS_ERR(vol_args)) { | |
2765 | ret = PTR_ERR(vol_args); | |
183860f6 AJ |
2766 | goto out; |
2767 | } | |
2768 | ||
58d7bbf8 | 2769 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
0b246afa | 2770 | mutex_lock(&fs_info->volume_mutex); |
2ff7e61e | 2771 | ret = btrfs_rm_device(fs_info, vol_args->name, 0); |
0b246afa | 2772 | mutex_unlock(&fs_info->volume_mutex); |
183860f6 | 2773 | |
ec95d491 | 2774 | if (!ret) |
0b246afa | 2775 | btrfs_info(fs_info, "disk deleted %s", vol_args->name); |
183860f6 | 2776 | kfree(vol_args); |
58d7bbf8 | 2777 | out: |
0b246afa | 2778 | atomic_set(&fs_info->mutually_exclusive_operation_running, 0); |
58d7bbf8 | 2779 | out_drop_write: |
4ac20c70 | 2780 | mnt_drop_write_file(file); |
58d7bbf8 | 2781 | |
f46b5a66 CH |
2782 | return ret; |
2783 | } | |
2784 | ||
2ff7e61e JM |
2785 | static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, |
2786 | void __user *arg) | |
475f6387 | 2787 | { |
027ed2f0 | 2788 | struct btrfs_ioctl_fs_info_args *fi_args; |
475f6387 | 2789 | struct btrfs_device *device; |
0b246afa | 2790 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; |
027ed2f0 | 2791 | int ret = 0; |
475f6387 | 2792 | |
027ed2f0 LZ |
2793 | fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); |
2794 | if (!fi_args) | |
2795 | return -ENOMEM; | |
2796 | ||
f7171750 | 2797 | mutex_lock(&fs_devices->device_list_mutex); |
027ed2f0 | 2798 | fi_args->num_devices = fs_devices->num_devices; |
0b246afa | 2799 | memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid)); |
475f6387 | 2800 | |
d7641a49 | 2801 | list_for_each_entry(device, &fs_devices->devices, dev_list) { |
027ed2f0 LZ |
2802 | if (device->devid > fi_args->max_id) |
2803 | fi_args->max_id = device->devid; | |
475f6387 JS |
2804 | } |
2805 | mutex_unlock(&fs_devices->device_list_mutex); | |
2806 | ||
0b246afa JM |
2807 | fi_args->nodesize = fs_info->super_copy->nodesize; |
2808 | fi_args->sectorsize = fs_info->super_copy->sectorsize; | |
2809 | fi_args->clone_alignment = fs_info->super_copy->sectorsize; | |
80a773fb | 2810 | |
027ed2f0 LZ |
2811 | if (copy_to_user(arg, fi_args, sizeof(*fi_args))) |
2812 | ret = -EFAULT; | |
475f6387 | 2813 | |
027ed2f0 LZ |
2814 | kfree(fi_args); |
2815 | return ret; | |
475f6387 JS |
2816 | } |
2817 | ||
2ff7e61e JM |
2818 | static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, |
2819 | void __user *arg) | |
475f6387 JS |
2820 | { |
2821 | struct btrfs_ioctl_dev_info_args *di_args; | |
2822 | struct btrfs_device *dev; | |
0b246afa | 2823 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; |
475f6387 JS |
2824 | int ret = 0; |
2825 | char *s_uuid = NULL; | |
475f6387 | 2826 | |
475f6387 JS |
2827 | di_args = memdup_user(arg, sizeof(*di_args)); |
2828 | if (IS_ERR(di_args)) | |
2829 | return PTR_ERR(di_args); | |
2830 | ||
dd5f9615 | 2831 | if (!btrfs_is_empty_uuid(di_args->uuid)) |
475f6387 JS |
2832 | s_uuid = di_args->uuid; |
2833 | ||
2834 | mutex_lock(&fs_devices->device_list_mutex); | |
0b246afa | 2835 | dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL); |
475f6387 JS |
2836 | |
2837 | if (!dev) { | |
2838 | ret = -ENODEV; | |
2839 | goto out; | |
2840 | } | |
2841 | ||
2842 | di_args->devid = dev->devid; | |
7cc8e58d MX |
2843 | di_args->bytes_used = btrfs_device_get_bytes_used(dev); |
2844 | di_args->total_bytes = btrfs_device_get_total_bytes(dev); | |
475f6387 | 2845 | memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); |
a27202fb | 2846 | if (dev->name) { |
606686ee JB |
2847 | struct rcu_string *name; |
2848 | ||
2849 | rcu_read_lock(); | |
2850 | name = rcu_dereference(dev->name); | |
2851 | strncpy(di_args->path, name->str, sizeof(di_args->path)); | |
2852 | rcu_read_unlock(); | |
a27202fb JM |
2853 | di_args->path[sizeof(di_args->path) - 1] = 0; |
2854 | } else { | |
99ba55ad | 2855 | di_args->path[0] = '\0'; |
a27202fb | 2856 | } |
475f6387 JS |
2857 | |
2858 | out: | |
55793c0d | 2859 | mutex_unlock(&fs_devices->device_list_mutex); |
475f6387 JS |
2860 | if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) |
2861 | ret = -EFAULT; | |
2862 | ||
2863 | kfree(di_args); | |
2864 | return ret; | |
2865 | } | |
2866 | ||
f4414602 | 2867 | static struct page *extent_same_get_page(struct inode *inode, pgoff_t index) |
416161db MF |
2868 | { |
2869 | struct page *page; | |
416161db | 2870 | |
416161db MF |
2871 | page = grab_cache_page(inode->i_mapping, index); |
2872 | if (!page) | |
31314002 | 2873 | return ERR_PTR(-ENOMEM); |
416161db MF |
2874 | |
2875 | if (!PageUptodate(page)) { | |
31314002 FM |
2876 | int ret; |
2877 | ||
2878 | ret = btrfs_readpage(NULL, page); | |
2879 | if (ret) | |
2880 | return ERR_PTR(ret); | |
416161db MF |
2881 | lock_page(page); |
2882 | if (!PageUptodate(page)) { | |
2883 | unlock_page(page); | |
09cbfeaf | 2884 | put_page(page); |
31314002 FM |
2885 | return ERR_PTR(-EIO); |
2886 | } | |
2887 | if (page->mapping != inode->i_mapping) { | |
2888 | unlock_page(page); | |
09cbfeaf | 2889 | put_page(page); |
31314002 | 2890 | return ERR_PTR(-EAGAIN); |
416161db MF |
2891 | } |
2892 | } | |
416161db MF |
2893 | |
2894 | return page; | |
2895 | } | |
2896 | ||
f4414602 MF |
2897 | static int gather_extent_pages(struct inode *inode, struct page **pages, |
2898 | int num_pages, u64 off) | |
2899 | { | |
2900 | int i; | |
09cbfeaf | 2901 | pgoff_t index = off >> PAGE_SHIFT; |
f4414602 MF |
2902 | |
2903 | for (i = 0; i < num_pages; i++) { | |
31314002 | 2904 | again: |
f4414602 | 2905 | pages[i] = extent_same_get_page(inode, index + i); |
31314002 FM |
2906 | if (IS_ERR(pages[i])) { |
2907 | int err = PTR_ERR(pages[i]); | |
2908 | ||
2909 | if (err == -EAGAIN) | |
2910 | goto again; | |
2911 | pages[i] = NULL; | |
2912 | return err; | |
2913 | } | |
f4414602 MF |
2914 | } |
2915 | return 0; | |
2916 | } | |
2917 | ||
e0bd70c6 FM |
2918 | static int lock_extent_range(struct inode *inode, u64 off, u64 len, |
2919 | bool retry_range_locking) | |
77fe20dc | 2920 | { |
e0bd70c6 FM |
2921 | /* |
2922 | * Do any pending delalloc/csum calculations on inode, one way or | |
2923 | * another, and lock file content. | |
2924 | * The locking order is: | |
2925 | * | |
2926 | * 1) pages | |
2927 | * 2) range in the inode's io tree | |
2928 | */ | |
77fe20dc MF |
2929 | while (1) { |
2930 | struct btrfs_ordered_extent *ordered; | |
2931 | lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); | |
2932 | ordered = btrfs_lookup_first_ordered_extent(inode, | |
2933 | off + len - 1); | |
ff5df9b8 FM |
2934 | if ((!ordered || |
2935 | ordered->file_offset + ordered->len <= off || | |
2936 | ordered->file_offset >= off + len) && | |
77fe20dc | 2937 | !test_range_bit(&BTRFS_I(inode)->io_tree, off, |
ff5df9b8 FM |
2938 | off + len - 1, EXTENT_DELALLOC, 0, NULL)) { |
2939 | if (ordered) | |
2940 | btrfs_put_ordered_extent(ordered); | |
77fe20dc | 2941 | break; |
ff5df9b8 | 2942 | } |
77fe20dc MF |
2943 | unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); |
2944 | if (ordered) | |
2945 | btrfs_put_ordered_extent(ordered); | |
e0bd70c6 FM |
2946 | if (!retry_range_locking) |
2947 | return -EAGAIN; | |
77fe20dc MF |
2948 | btrfs_wait_ordered_range(inode, off, len); |
2949 | } | |
e0bd70c6 | 2950 | return 0; |
77fe20dc MF |
2951 | } |
2952 | ||
f4414602 | 2953 | static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2) |
416161db | 2954 | { |
5955102c AV |
2955 | inode_unlock(inode1); |
2956 | inode_unlock(inode2); | |
416161db MF |
2957 | } |
2958 | ||
f4414602 MF |
2959 | static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2) |
2960 | { | |
2961 | if (inode1 < inode2) | |
2962 | swap(inode1, inode2); | |
2963 | ||
5955102c AV |
2964 | inode_lock_nested(inode1, I_MUTEX_PARENT); |
2965 | inode_lock_nested(inode2, I_MUTEX_CHILD); | |
f4414602 MF |
2966 | } |
2967 | ||
2968 | static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, | |
2969 | struct inode *inode2, u64 loff2, u64 len) | |
2970 | { | |
2971 | unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); | |
2972 | unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); | |
2973 | } | |
2974 | ||
e0bd70c6 FM |
2975 | static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1, |
2976 | struct inode *inode2, u64 loff2, u64 len, | |
2977 | bool retry_range_locking) | |
416161db | 2978 | { |
e0bd70c6 FM |
2979 | int ret; |
2980 | ||
416161db MF |
2981 | if (inode1 < inode2) { |
2982 | swap(inode1, inode2); | |
2983 | swap(loff1, loff2); | |
2984 | } | |
e0bd70c6 FM |
2985 | ret = lock_extent_range(inode1, loff1, len, retry_range_locking); |
2986 | if (ret) | |
2987 | return ret; | |
2988 | ret = lock_extent_range(inode2, loff2, len, retry_range_locking); | |
2989 | if (ret) | |
2990 | unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, | |
2991 | loff1 + len - 1); | |
2992 | return ret; | |
f4414602 MF |
2993 | } |
2994 | ||
2995 | struct cmp_pages { | |
2996 | int num_pages; | |
2997 | struct page **src_pages; | |
2998 | struct page **dst_pages; | |
2999 | }; | |
3000 | ||
3001 | static void btrfs_cmp_data_free(struct cmp_pages *cmp) | |
3002 | { | |
3003 | int i; | |
3004 | struct page *pg; | |
3005 | ||
3006 | for (i = 0; i < cmp->num_pages; i++) { | |
3007 | pg = cmp->src_pages[i]; | |
e0bd70c6 FM |
3008 | if (pg) { |
3009 | unlock_page(pg); | |
09cbfeaf | 3010 | put_page(pg); |
e0bd70c6 | 3011 | } |
f4414602 | 3012 | pg = cmp->dst_pages[i]; |
e0bd70c6 FM |
3013 | if (pg) { |
3014 | unlock_page(pg); | |
09cbfeaf | 3015 | put_page(pg); |
e0bd70c6 | 3016 | } |
f4414602 MF |
3017 | } |
3018 | kfree(cmp->src_pages); | |
3019 | kfree(cmp->dst_pages); | |
3020 | } | |
3021 | ||
3022 | static int btrfs_cmp_data_prepare(struct inode *src, u64 loff, | |
3023 | struct inode *dst, u64 dst_loff, | |
3024 | u64 len, struct cmp_pages *cmp) | |
3025 | { | |
3026 | int ret; | |
09cbfeaf | 3027 | int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT; |
f4414602 MF |
3028 | struct page **src_pgarr, **dst_pgarr; |
3029 | ||
3030 | /* | |
3031 | * We must gather up all the pages before we initiate our | |
3032 | * extent locking. We use an array for the page pointers. Size | |
3033 | * of the array is bounded by len, which is in turn bounded by | |
3034 | * BTRFS_MAX_DEDUPE_LEN. | |
3035 | */ | |
66722f7c DS |
3036 | src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL); |
3037 | dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL); | |
f4414602 MF |
3038 | if (!src_pgarr || !dst_pgarr) { |
3039 | kfree(src_pgarr); | |
3040 | kfree(dst_pgarr); | |
3041 | return -ENOMEM; | |
416161db | 3042 | } |
f4414602 MF |
3043 | cmp->num_pages = num_pages; |
3044 | cmp->src_pages = src_pgarr; | |
3045 | cmp->dst_pages = dst_pgarr; | |
3046 | ||
3047 | ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff); | |
3048 | if (ret) | |
3049 | goto out; | |
3050 | ||
3051 | ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff); | |
3052 | ||
3053 | out: | |
3054 | if (ret) | |
3055 | btrfs_cmp_data_free(cmp); | |
3056 | return 0; | |
416161db MF |
3057 | } |
3058 | ||
3059 | static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst, | |
f4414602 | 3060 | u64 dst_loff, u64 len, struct cmp_pages *cmp) |
416161db MF |
3061 | { |
3062 | int ret = 0; | |
f4414602 | 3063 | int i; |
416161db | 3064 | struct page *src_page, *dst_page; |
09cbfeaf | 3065 | unsigned int cmp_len = PAGE_SIZE; |
416161db MF |
3066 | void *addr, *dst_addr; |
3067 | ||
f4414602 | 3068 | i = 0; |
416161db | 3069 | while (len) { |
09cbfeaf | 3070 | if (len < PAGE_SIZE) |
416161db MF |
3071 | cmp_len = len; |
3072 | ||
f4414602 MF |
3073 | BUG_ON(i >= cmp->num_pages); |
3074 | ||
3075 | src_page = cmp->src_pages[i]; | |
3076 | dst_page = cmp->dst_pages[i]; | |
e0bd70c6 FM |
3077 | ASSERT(PageLocked(src_page)); |
3078 | ASSERT(PageLocked(dst_page)); | |
f4414602 | 3079 | |
416161db MF |
3080 | addr = kmap_atomic(src_page); |
3081 | dst_addr = kmap_atomic(dst_page); | |
3082 | ||
3083 | flush_dcache_page(src_page); | |
3084 | flush_dcache_page(dst_page); | |
3085 | ||
3086 | if (memcmp(addr, dst_addr, cmp_len)) | |
2b3909f8 | 3087 | ret = -EBADE; |
416161db MF |
3088 | |
3089 | kunmap_atomic(addr); | |
3090 | kunmap_atomic(dst_addr); | |
416161db MF |
3091 | |
3092 | if (ret) | |
3093 | break; | |
3094 | ||
416161db | 3095 | len -= cmp_len; |
f4414602 | 3096 | i++; |
416161db MF |
3097 | } |
3098 | ||
3099 | return ret; | |
3100 | } | |
3101 | ||
e1d227a4 MF |
3102 | static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen, |
3103 | u64 olen) | |
416161db | 3104 | { |
e1d227a4 | 3105 | u64 len = *plen; |
416161db MF |
3106 | u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; |
3107 | ||
e1d227a4 | 3108 | if (off + olen > inode->i_size || off + olen < off) |
416161db | 3109 | return -EINVAL; |
e1d227a4 MF |
3110 | |
3111 | /* if we extend to eof, continue to block boundary */ | |
3112 | if (off + len == inode->i_size) | |
3113 | *plen = len = ALIGN(inode->i_size, bs) - off; | |
3114 | ||
416161db MF |
3115 | /* Check that we are block aligned - btrfs_clone() requires this */ |
3116 | if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) | |
3117 | return -EINVAL; | |
3118 | ||
3119 | return 0; | |
3120 | } | |
3121 | ||
e1d227a4 | 3122 | static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, |
416161db MF |
3123 | struct inode *dst, u64 dst_loff) |
3124 | { | |
3125 | int ret; | |
e1d227a4 | 3126 | u64 len = olen; |
f4414602 | 3127 | struct cmp_pages cmp; |
0efa9f48 MF |
3128 | int same_inode = 0; |
3129 | u64 same_lock_start = 0; | |
3130 | u64 same_lock_len = 0; | |
416161db | 3131 | |
416161db | 3132 | if (src == dst) |
0efa9f48 | 3133 | same_inode = 1; |
416161db | 3134 | |
113e8283 FM |
3135 | if (len == 0) |
3136 | return 0; | |
3137 | ||
0efa9f48 | 3138 | if (same_inode) { |
5955102c | 3139 | inode_lock(src); |
416161db | 3140 | |
0efa9f48 | 3141 | ret = extent_same_check_offsets(src, loff, &len, olen); |
f4dfe687 FM |
3142 | if (ret) |
3143 | goto out_unlock; | |
3144 | ret = extent_same_check_offsets(src, dst_loff, &len, olen); | |
0efa9f48 MF |
3145 | if (ret) |
3146 | goto out_unlock; | |
416161db | 3147 | |
0efa9f48 MF |
3148 | /* |
3149 | * Single inode case wants the same checks, except we | |
3150 | * don't want our length pushed out past i_size as | |
3151 | * comparing that data range makes no sense. | |
3152 | * | |
3153 | * extent_same_check_offsets() will do this for an | |
3154 | * unaligned length at i_size, so catch it here and | |
3155 | * reject the request. | |
3156 | * | |
3157 | * This effectively means we require aligned extents | |
3158 | * for the single-inode case, whereas the other cases | |
3159 | * allow an unaligned length so long as it ends at | |
3160 | * i_size. | |
3161 | */ | |
3162 | if (len != olen) { | |
3163 | ret = -EINVAL; | |
3164 | goto out_unlock; | |
3165 | } | |
3166 | ||
3167 | /* Check for overlapping ranges */ | |
3168 | if (dst_loff + len > loff && dst_loff < loff + len) { | |
3169 | ret = -EINVAL; | |
3170 | goto out_unlock; | |
3171 | } | |
3172 | ||
3173 | same_lock_start = min_t(u64, loff, dst_loff); | |
3174 | same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start; | |
3175 | } else { | |
3176 | btrfs_double_inode_lock(src, dst); | |
3177 | ||
3178 | ret = extent_same_check_offsets(src, loff, &len, olen); | |
3179 | if (ret) | |
3180 | goto out_unlock; | |
3181 | ||
3182 | ret = extent_same_check_offsets(dst, dst_loff, &len, olen); | |
3183 | if (ret) | |
3184 | goto out_unlock; | |
3185 | } | |
416161db MF |
3186 | |
3187 | /* don't make the dst file partly checksummed */ | |
3188 | if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != | |
3189 | (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { | |
3190 | ret = -EINVAL; | |
3191 | goto out_unlock; | |
3192 | } | |
3193 | ||
e0bd70c6 | 3194 | again: |
f4414602 MF |
3195 | ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp); |
3196 | if (ret) | |
3197 | goto out_unlock; | |
3198 | ||
0efa9f48 | 3199 | if (same_inode) |
e0bd70c6 FM |
3200 | ret = lock_extent_range(src, same_lock_start, same_lock_len, |
3201 | false); | |
0efa9f48 | 3202 | else |
e0bd70c6 FM |
3203 | ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len, |
3204 | false); | |
3205 | /* | |
3206 | * If one of the inodes has dirty pages in the respective range or | |
3207 | * ordered extents, we need to flush dellaloc and wait for all ordered | |
3208 | * extents in the range. We must unlock the pages and the ranges in the | |
3209 | * io trees to avoid deadlocks when flushing delalloc (requires locking | |
3210 | * pages) and when waiting for ordered extents to complete (they require | |
3211 | * range locking). | |
3212 | */ | |
3213 | if (ret == -EAGAIN) { | |
3214 | /* | |
3215 | * Ranges in the io trees already unlocked. Now unlock all | |
3216 | * pages before waiting for all IO to complete. | |
3217 | */ | |
3218 | btrfs_cmp_data_free(&cmp); | |
3219 | if (same_inode) { | |
3220 | btrfs_wait_ordered_range(src, same_lock_start, | |
3221 | same_lock_len); | |
3222 | } else { | |
3223 | btrfs_wait_ordered_range(src, loff, len); | |
3224 | btrfs_wait_ordered_range(dst, dst_loff, len); | |
3225 | } | |
3226 | goto again; | |
3227 | } | |
3228 | ASSERT(ret == 0); | |
3229 | if (WARN_ON(ret)) { | |
3230 | /* ranges in the io trees already unlocked */ | |
3231 | btrfs_cmp_data_free(&cmp); | |
3232 | return ret; | |
3233 | } | |
f4414602 | 3234 | |
207910dd | 3235 | /* pass original length for comparison so we stay within i_size */ |
f4414602 | 3236 | ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp); |
416161db | 3237 | if (ret == 0) |
1c919a5e | 3238 | ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1); |
416161db | 3239 | |
0efa9f48 MF |
3240 | if (same_inode) |
3241 | unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start, | |
3242 | same_lock_start + same_lock_len - 1); | |
3243 | else | |
3244 | btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); | |
f4414602 MF |
3245 | |
3246 | btrfs_cmp_data_free(&cmp); | |
416161db | 3247 | out_unlock: |
0efa9f48 | 3248 | if (same_inode) |
5955102c | 3249 | inode_unlock(src); |
0efa9f48 MF |
3250 | else |
3251 | btrfs_double_inode_unlock(src, dst); | |
416161db MF |
3252 | |
3253 | return ret; | |
3254 | } | |
3255 | ||
ee22184b | 3256 | #define BTRFS_MAX_DEDUPE_LEN SZ_16M |
416161db | 3257 | |
2b3909f8 DW |
3258 | ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen, |
3259 | struct file *dst_file, u64 dst_loff) | |
416161db | 3260 | { |
2b3909f8 DW |
3261 | struct inode *src = file_inode(src_file); |
3262 | struct inode *dst = file_inode(dst_file); | |
416161db | 3263 | u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; |
2b3909f8 | 3264 | ssize_t res; |
416161db | 3265 | |
2b3909f8 DW |
3266 | if (olen > BTRFS_MAX_DEDUPE_LEN) |
3267 | olen = BTRFS_MAX_DEDUPE_LEN; | |
416161db | 3268 | |
09cbfeaf | 3269 | if (WARN_ON_ONCE(bs < PAGE_SIZE)) { |
416161db MF |
3270 | /* |
3271 | * Btrfs does not support blocksize < page_size. As a | |
3272 | * result, btrfs_cmp_data() won't correctly handle | |
3273 | * this situation without an update. | |
3274 | */ | |
2b3909f8 | 3275 | return -EINVAL; |
416161db MF |
3276 | } |
3277 | ||
2b3909f8 DW |
3278 | res = btrfs_extent_same(src, loff, olen, dst, dst_loff); |
3279 | if (res) | |
3280 | return res; | |
3281 | return olen; | |
416161db MF |
3282 | } |
3283 | ||
f82a9901 FM |
3284 | static int clone_finish_inode_update(struct btrfs_trans_handle *trans, |
3285 | struct inode *inode, | |
3286 | u64 endoff, | |
3287 | const u64 destoff, | |
1c919a5e MF |
3288 | const u64 olen, |
3289 | int no_time_update) | |
f82a9901 FM |
3290 | { |
3291 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
3292 | int ret; | |
3293 | ||
3294 | inode_inc_iversion(inode); | |
1c919a5e | 3295 | if (!no_time_update) |
c2050a45 | 3296 | inode->i_mtime = inode->i_ctime = current_time(inode); |
f82a9901 FM |
3297 | /* |
3298 | * We round up to the block size at eof when determining which | |
3299 | * extents to clone above, but shouldn't round up the file size. | |
3300 | */ | |
3301 | if (endoff > destoff + olen) | |
3302 | endoff = destoff + olen; | |
3303 | if (endoff > inode->i_size) | |
3304 | btrfs_i_size_write(inode, endoff); | |
3305 | ||
3306 | ret = btrfs_update_inode(trans, root, inode); | |
3307 | if (ret) { | |
66642832 | 3308 | btrfs_abort_transaction(trans, ret); |
3a45bb20 | 3309 | btrfs_end_transaction(trans); |
f82a9901 FM |
3310 | goto out; |
3311 | } | |
3a45bb20 | 3312 | ret = btrfs_end_transaction(trans); |
f82a9901 FM |
3313 | out: |
3314 | return ret; | |
3315 | } | |
3316 | ||
7ffbb598 FM |
3317 | static void clone_update_extent_map(struct inode *inode, |
3318 | const struct btrfs_trans_handle *trans, | |
3319 | const struct btrfs_path *path, | |
7ffbb598 FM |
3320 | const u64 hole_offset, |
3321 | const u64 hole_len) | |
3322 | { | |
3323 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
3324 | struct extent_map *em; | |
3325 | int ret; | |
3326 | ||
3327 | em = alloc_extent_map(); | |
3328 | if (!em) { | |
3329 | set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, | |
3330 | &BTRFS_I(inode)->runtime_flags); | |
3331 | return; | |
3332 | } | |
3333 | ||
14f59796 FM |
3334 | if (path) { |
3335 | struct btrfs_file_extent_item *fi; | |
3336 | ||
3337 | fi = btrfs_item_ptr(path->nodes[0], path->slots[0], | |
3338 | struct btrfs_file_extent_item); | |
7ffbb598 FM |
3339 | btrfs_extent_item_to_extent_map(inode, path, fi, false, em); |
3340 | em->generation = -1; | |
3341 | if (btrfs_file_extent_type(path->nodes[0], fi) == | |
3342 | BTRFS_FILE_EXTENT_INLINE) | |
3343 | set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, | |
3344 | &BTRFS_I(inode)->runtime_flags); | |
3345 | } else { | |
3346 | em->start = hole_offset; | |
3347 | em->len = hole_len; | |
3348 | em->ram_bytes = em->len; | |
3349 | em->orig_start = hole_offset; | |
3350 | em->block_start = EXTENT_MAP_HOLE; | |
3351 | em->block_len = 0; | |
3352 | em->orig_block_len = 0; | |
3353 | em->compress_type = BTRFS_COMPRESS_NONE; | |
3354 | em->generation = trans->transid; | |
3355 | } | |
3356 | ||
3357 | while (1) { | |
3358 | write_lock(&em_tree->lock); | |
3359 | ret = add_extent_mapping(em_tree, em, 1); | |
3360 | write_unlock(&em_tree->lock); | |
3361 | if (ret != -EEXIST) { | |
3362 | free_extent_map(em); | |
3363 | break; | |
3364 | } | |
3365 | btrfs_drop_extent_cache(inode, em->start, | |
3366 | em->start + em->len - 1, 0); | |
3367 | } | |
3368 | ||
ee39b432 | 3369 | if (ret) |
7ffbb598 FM |
3370 | set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, |
3371 | &BTRFS_I(inode)->runtime_flags); | |
3372 | } | |
3373 | ||
8039d87d FM |
3374 | /* |
3375 | * Make sure we do not end up inserting an inline extent into a file that has | |
3376 | * already other (non-inline) extents. If a file has an inline extent it can | |
3377 | * not have any other extents and the (single) inline extent must start at the | |
3378 | * file offset 0. Failing to respect these rules will lead to file corruption, | |
3379 | * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc | |
3380 | * | |
3381 | * We can have extents that have been already written to disk or we can have | |
3382 | * dirty ranges still in delalloc, in which case the extent maps and items are | |
3383 | * created only when we run delalloc, and the delalloc ranges might fall outside | |
3384 | * the range we are currently locking in the inode's io tree. So we check the | |
3385 | * inode's i_size because of that (i_size updates are done while holding the | |
3386 | * i_mutex, which we are holding here). | |
3387 | * We also check to see if the inode has a size not greater than "datal" but has | |
3388 | * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are | |
3389 | * protected against such concurrent fallocate calls by the i_mutex). | |
3390 | * | |
3391 | * If the file has no extents but a size greater than datal, do not allow the | |
3392 | * copy because we would need turn the inline extent into a non-inline one (even | |
3393 | * with NO_HOLES enabled). If we find our destination inode only has one inline | |
3394 | * extent, just overwrite it with the source inline extent if its size is less | |
3395 | * than the source extent's size, or we could copy the source inline extent's | |
3396 | * data into the destination inode's inline extent if the later is greater then | |
3397 | * the former. | |
3398 | */ | |
3399 | static int clone_copy_inline_extent(struct inode *src, | |
3400 | struct inode *dst, | |
3401 | struct btrfs_trans_handle *trans, | |
3402 | struct btrfs_path *path, | |
3403 | struct btrfs_key *new_key, | |
3404 | const u64 drop_start, | |
3405 | const u64 datal, | |
3406 | const u64 skip, | |
3407 | const u64 size, | |
3408 | char *inline_data) | |
3409 | { | |
0b246afa | 3410 | struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb); |
8039d87d FM |
3411 | struct btrfs_root *root = BTRFS_I(dst)->root; |
3412 | const u64 aligned_end = ALIGN(new_key->offset + datal, | |
0b246afa | 3413 | fs_info->sectorsize); |
8039d87d FM |
3414 | int ret; |
3415 | struct btrfs_key key; | |
3416 | ||
3417 | if (new_key->offset > 0) | |
3418 | return -EOPNOTSUPP; | |
3419 | ||
3420 | key.objectid = btrfs_ino(dst); | |
3421 | key.type = BTRFS_EXTENT_DATA_KEY; | |
3422 | key.offset = 0; | |
3423 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
3424 | if (ret < 0) { | |
3425 | return ret; | |
3426 | } else if (ret > 0) { | |
3427 | if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { | |
3428 | ret = btrfs_next_leaf(root, path); | |
3429 | if (ret < 0) | |
3430 | return ret; | |
3431 | else if (ret > 0) | |
3432 | goto copy_inline_extent; | |
3433 | } | |
3434 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); | |
3435 | if (key.objectid == btrfs_ino(dst) && | |
3436 | key.type == BTRFS_EXTENT_DATA_KEY) { | |
3437 | ASSERT(key.offset > 0); | |
3438 | return -EOPNOTSUPP; | |
3439 | } | |
3440 | } else if (i_size_read(dst) <= datal) { | |
3441 | struct btrfs_file_extent_item *ei; | |
3442 | u64 ext_len; | |
3443 | ||
3444 | /* | |
3445 | * If the file size is <= datal, make sure there are no other | |
3446 | * extents following (can happen do to an fallocate call with | |
3447 | * the flag FALLOC_FL_KEEP_SIZE). | |
3448 | */ | |
3449 | ei = btrfs_item_ptr(path->nodes[0], path->slots[0], | |
3450 | struct btrfs_file_extent_item); | |
3451 | /* | |
3452 | * If it's an inline extent, it can not have other extents | |
3453 | * following it. | |
3454 | */ | |
3455 | if (btrfs_file_extent_type(path->nodes[0], ei) == | |
3456 | BTRFS_FILE_EXTENT_INLINE) | |
3457 | goto copy_inline_extent; | |
3458 | ||
3459 | ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); | |
3460 | if (ext_len > aligned_end) | |
3461 | return -EOPNOTSUPP; | |
3462 | ||
3463 | ret = btrfs_next_item(root, path); | |
3464 | if (ret < 0) { | |
3465 | return ret; | |
3466 | } else if (ret == 0) { | |
3467 | btrfs_item_key_to_cpu(path->nodes[0], &key, | |
3468 | path->slots[0]); | |
3469 | if (key.objectid == btrfs_ino(dst) && | |
3470 | key.type == BTRFS_EXTENT_DATA_KEY) | |
3471 | return -EOPNOTSUPP; | |
3472 | } | |
3473 | } | |
3474 | ||
3475 | copy_inline_extent: | |
3476 | /* | |
3477 | * We have no extent items, or we have an extent at offset 0 which may | |
3478 | * or may not be inlined. All these cases are dealt the same way. | |
3479 | */ | |
3480 | if (i_size_read(dst) > datal) { | |
3481 | /* | |
3482 | * If the destination inode has an inline extent... | |
3483 | * This would require copying the data from the source inline | |
3484 | * extent into the beginning of the destination's inline extent. | |
3485 | * But this is really complex, both extents can be compressed | |
3486 | * or just one of them, which would require decompressing and | |
3487 | * re-compressing data (which could increase the new compressed | |
3488 | * size, not allowing the compressed data to fit anymore in an | |
3489 | * inline extent). | |
3490 | * So just don't support this case for now (it should be rare, | |
3491 | * we are not really saving space when cloning inline extents). | |
3492 | */ | |
3493 | return -EOPNOTSUPP; | |
3494 | } | |
3495 | ||
3496 | btrfs_release_path(path); | |
3497 | ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); | |
3498 | if (ret) | |
3499 | return ret; | |
3500 | ret = btrfs_insert_empty_item(trans, root, path, new_key, size); | |
3501 | if (ret) | |
3502 | return ret; | |
3503 | ||
3504 | if (skip) { | |
3505 | const u32 start = btrfs_file_extent_calc_inline_size(0); | |
3506 | ||
3507 | memmove(inline_data + start, inline_data + start + skip, datal); | |
3508 | } | |
3509 | ||
3510 | write_extent_buffer(path->nodes[0], inline_data, | |
3511 | btrfs_item_ptr_offset(path->nodes[0], | |
3512 | path->slots[0]), | |
3513 | size); | |
3514 | inode_add_bytes(dst, datal); | |
3515 | ||
3516 | return 0; | |
3517 | } | |
3518 | ||
32b7c687 MF |
3519 | /** |
3520 | * btrfs_clone() - clone a range from inode file to another | |
3521 | * | |
3522 | * @src: Inode to clone from | |
3523 | * @inode: Inode to clone to | |
3524 | * @off: Offset within source to start clone from | |
3525 | * @olen: Original length, passed by user, of range to clone | |
1c919a5e | 3526 | * @olen_aligned: Block-aligned value of olen |
32b7c687 | 3527 | * @destoff: Offset within @inode to start clone |
1c919a5e | 3528 | * @no_time_update: Whether to update mtime/ctime on the target inode |
32b7c687 MF |
3529 | */ |
3530 | static int btrfs_clone(struct inode *src, struct inode *inode, | |
f82a9901 | 3531 | const u64 off, const u64 olen, const u64 olen_aligned, |
1c919a5e | 3532 | const u64 destoff, int no_time_update) |
f46b5a66 | 3533 | { |
0b246afa | 3534 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
f46b5a66 | 3535 | struct btrfs_root *root = BTRFS_I(inode)->root; |
32b7c687 | 3536 | struct btrfs_path *path = NULL; |
f46b5a66 | 3537 | struct extent_buffer *leaf; |
32b7c687 MF |
3538 | struct btrfs_trans_handle *trans; |
3539 | char *buf = NULL; | |
ae01a0ab | 3540 | struct btrfs_key key; |
f46b5a66 CH |
3541 | u32 nritems; |
3542 | int slot; | |
ae01a0ab | 3543 | int ret; |
f82a9901 | 3544 | const u64 len = olen_aligned; |
f82a9901 | 3545 | u64 last_dest_end = destoff; |
ae01a0ab YZ |
3546 | |
3547 | ret = -ENOMEM; | |
0b246afa | 3548 | buf = kmalloc(fs_info->nodesize, GFP_KERNEL | __GFP_NOWARN); |
15351955 | 3549 | if (!buf) { |
0b246afa | 3550 | buf = vmalloc(fs_info->nodesize); |
15351955 DS |
3551 | if (!buf) |
3552 | return ret; | |
3553 | } | |
ae01a0ab YZ |
3554 | |
3555 | path = btrfs_alloc_path(); | |
3556 | if (!path) { | |
15351955 | 3557 | kvfree(buf); |
32b7c687 | 3558 | return ret; |
d525e8ab LZ |
3559 | } |
3560 | ||
e4058b54 | 3561 | path->reada = READA_FORWARD; |
c5c9cd4d | 3562 | /* clone data */ |
33345d01 | 3563 | key.objectid = btrfs_ino(src); |
ae01a0ab | 3564 | key.type = BTRFS_EXTENT_DATA_KEY; |
2c463823 | 3565 | key.offset = off; |
f46b5a66 CH |
3566 | |
3567 | while (1) { | |
de249e66 | 3568 | u64 next_key_min_offset = key.offset + 1; |
df858e76 | 3569 | |
f46b5a66 CH |
3570 | /* |
3571 | * note the key will change type as we walk through the | |
3572 | * tree. | |
3573 | */ | |
e4355f34 | 3574 | path->leave_spinning = 1; |
362a20c5 DS |
3575 | ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, |
3576 | 0, 0); | |
f46b5a66 CH |
3577 | if (ret < 0) |
3578 | goto out; | |
2c463823 FM |
3579 | /* |
3580 | * First search, if no extent item that starts at offset off was | |
3581 | * found but the previous item is an extent item, it's possible | |
3582 | * it might overlap our target range, therefore process it. | |
3583 | */ | |
3584 | if (key.offset == off && ret > 0 && path->slots[0] > 0) { | |
3585 | btrfs_item_key_to_cpu(path->nodes[0], &key, | |
3586 | path->slots[0] - 1); | |
3587 | if (key.type == BTRFS_EXTENT_DATA_KEY) | |
3588 | path->slots[0]--; | |
3589 | } | |
f46b5a66 | 3590 | |
ae01a0ab | 3591 | nritems = btrfs_header_nritems(path->nodes[0]); |
e4355f34 | 3592 | process_slot: |
ae01a0ab | 3593 | if (path->slots[0] >= nritems) { |
362a20c5 | 3594 | ret = btrfs_next_leaf(BTRFS_I(src)->root, path); |
f46b5a66 CH |
3595 | if (ret < 0) |
3596 | goto out; | |
3597 | if (ret > 0) | |
3598 | break; | |
ae01a0ab | 3599 | nritems = btrfs_header_nritems(path->nodes[0]); |
f46b5a66 CH |
3600 | } |
3601 | leaf = path->nodes[0]; | |
3602 | slot = path->slots[0]; | |
f46b5a66 | 3603 | |
ae01a0ab | 3604 | btrfs_item_key_to_cpu(leaf, &key, slot); |
962a298f | 3605 | if (key.type > BTRFS_EXTENT_DATA_KEY || |
33345d01 | 3606 | key.objectid != btrfs_ino(src)) |
f46b5a66 CH |
3607 | break; |
3608 | ||
962a298f | 3609 | if (key.type == BTRFS_EXTENT_DATA_KEY) { |
c5c9cd4d SW |
3610 | struct btrfs_file_extent_item *extent; |
3611 | int type; | |
31840ae1 ZY |
3612 | u32 size; |
3613 | struct btrfs_key new_key; | |
c5c9cd4d SW |
3614 | u64 disko = 0, diskl = 0; |
3615 | u64 datao = 0, datal = 0; | |
3616 | u8 comp; | |
f82a9901 | 3617 | u64 drop_start; |
31840ae1 | 3618 | |
c5c9cd4d SW |
3619 | extent = btrfs_item_ptr(leaf, slot, |
3620 | struct btrfs_file_extent_item); | |
3621 | comp = btrfs_file_extent_compression(leaf, extent); | |
3622 | type = btrfs_file_extent_type(leaf, extent); | |
c8a894d7 CM |
3623 | if (type == BTRFS_FILE_EXTENT_REG || |
3624 | type == BTRFS_FILE_EXTENT_PREALLOC) { | |
d397712b CM |
3625 | disko = btrfs_file_extent_disk_bytenr(leaf, |
3626 | extent); | |
3627 | diskl = btrfs_file_extent_disk_num_bytes(leaf, | |
3628 | extent); | |
c5c9cd4d | 3629 | datao = btrfs_file_extent_offset(leaf, extent); |
d397712b CM |
3630 | datal = btrfs_file_extent_num_bytes(leaf, |
3631 | extent); | |
c5c9cd4d SW |
3632 | } else if (type == BTRFS_FILE_EXTENT_INLINE) { |
3633 | /* take upper bound, may be compressed */ | |
3634 | datal = btrfs_file_extent_ram_bytes(leaf, | |
3635 | extent); | |
3636 | } | |
31840ae1 | 3637 | |
2c463823 FM |
3638 | /* |
3639 | * The first search might have left us at an extent | |
3640 | * item that ends before our target range's start, can | |
3641 | * happen if we have holes and NO_HOLES feature enabled. | |
3642 | */ | |
3643 | if (key.offset + datal <= off) { | |
e4355f34 FDBM |
3644 | path->slots[0]++; |
3645 | goto process_slot; | |
2c463823 FM |
3646 | } else if (key.offset >= off + len) { |
3647 | break; | |
e4355f34 | 3648 | } |
df858e76 | 3649 | next_key_min_offset = key.offset + datal; |
e4355f34 FDBM |
3650 | size = btrfs_item_size_nr(leaf, slot); |
3651 | read_extent_buffer(leaf, buf, | |
3652 | btrfs_item_ptr_offset(leaf, slot), | |
3653 | size); | |
3654 | ||
3655 | btrfs_release_path(path); | |
3656 | path->leave_spinning = 0; | |
c5c9cd4d | 3657 | |
31840ae1 | 3658 | memcpy(&new_key, &key, sizeof(new_key)); |
33345d01 | 3659 | new_key.objectid = btrfs_ino(inode); |
4d728ec7 LZ |
3660 | if (off <= key.offset) |
3661 | new_key.offset = key.offset + destoff - off; | |
3662 | else | |
3663 | new_key.offset = destoff; | |
31840ae1 | 3664 | |
f82a9901 FM |
3665 | /* |
3666 | * Deal with a hole that doesn't have an extent item | |
3667 | * that represents it (NO_HOLES feature enabled). | |
3668 | * This hole is either in the middle of the cloning | |
3669 | * range or at the beginning (fully overlaps it or | |
3670 | * partially overlaps it). | |
3671 | */ | |
3672 | if (new_key.offset != last_dest_end) | |
3673 | drop_start = last_dest_end; | |
3674 | else | |
3675 | drop_start = new_key.offset; | |
3676 | ||
b6f3409b SW |
3677 | /* |
3678 | * 1 - adjusting old extent (we may have to split it) | |
3679 | * 1 - add new extent | |
3680 | * 1 - inode update | |
3681 | */ | |
3682 | trans = btrfs_start_transaction(root, 3); | |
a22285a6 YZ |
3683 | if (IS_ERR(trans)) { |
3684 | ret = PTR_ERR(trans); | |
3685 | goto out; | |
3686 | } | |
3687 | ||
c8a894d7 CM |
3688 | if (type == BTRFS_FILE_EXTENT_REG || |
3689 | type == BTRFS_FILE_EXTENT_PREALLOC) { | |
d72c0842 LZ |
3690 | /* |
3691 | * a | --- range to clone ---| b | |
3692 | * | ------------- extent ------------- | | |
3693 | */ | |
3694 | ||
93915584 | 3695 | /* subtract range b */ |
d72c0842 LZ |
3696 | if (key.offset + datal > off + len) |
3697 | datal = off + len - key.offset; | |
3698 | ||
93915584 | 3699 | /* subtract range a */ |
a22285a6 YZ |
3700 | if (off > key.offset) { |
3701 | datao += off - key.offset; | |
3702 | datal -= off - key.offset; | |
3703 | } | |
3704 | ||
5dc562c5 | 3705 | ret = btrfs_drop_extents(trans, root, inode, |
f82a9901 | 3706 | drop_start, |
a22285a6 | 3707 | new_key.offset + datal, |
2671485d | 3708 | 1); |
79787eaa | 3709 | if (ret) { |
3f9e3df8 | 3710 | if (ret != -EOPNOTSUPP) |
00fdf13a | 3711 | btrfs_abort_transaction(trans, |
66642832 | 3712 | ret); |
3a45bb20 | 3713 | btrfs_end_transaction(trans); |
79787eaa JM |
3714 | goto out; |
3715 | } | |
a22285a6 | 3716 | |
c5c9cd4d SW |
3717 | ret = btrfs_insert_empty_item(trans, root, path, |
3718 | &new_key, size); | |
79787eaa | 3719 | if (ret) { |
66642832 | 3720 | btrfs_abort_transaction(trans, ret); |
3a45bb20 | 3721 | btrfs_end_transaction(trans); |
79787eaa JM |
3722 | goto out; |
3723 | } | |
c5c9cd4d SW |
3724 | |
3725 | leaf = path->nodes[0]; | |
3726 | slot = path->slots[0]; | |
3727 | write_extent_buffer(leaf, buf, | |
31840ae1 ZY |
3728 | btrfs_item_ptr_offset(leaf, slot), |
3729 | size); | |
ae01a0ab | 3730 | |
c5c9cd4d | 3731 | extent = btrfs_item_ptr(leaf, slot, |
f46b5a66 | 3732 | struct btrfs_file_extent_item); |
c5c9cd4d | 3733 | |
c5c9cd4d SW |
3734 | /* disko == 0 means it's a hole */ |
3735 | if (!disko) | |
3736 | datao = 0; | |
c5c9cd4d SW |
3737 | |
3738 | btrfs_set_file_extent_offset(leaf, extent, | |
3739 | datao); | |
3740 | btrfs_set_file_extent_num_bytes(leaf, extent, | |
3741 | datal); | |
fcebe456 | 3742 | |
c5c9cd4d SW |
3743 | if (disko) { |
3744 | inode_add_bytes(inode, datal); | |
2ff7e61e JM |
3745 | ret = btrfs_inc_extent_ref(trans, |
3746 | fs_info, | |
5d4f98a2 YZ |
3747 | disko, diskl, 0, |
3748 | root->root_key.objectid, | |
33345d01 | 3749 | btrfs_ino(inode), |
b06c4bf5 | 3750 | new_key.offset - datao); |
79787eaa JM |
3751 | if (ret) { |
3752 | btrfs_abort_transaction(trans, | |
79787eaa | 3753 | ret); |
3a45bb20 | 3754 | btrfs_end_transaction(trans); |
79787eaa JM |
3755 | goto out; |
3756 | ||
3757 | } | |
f46b5a66 | 3758 | } |
c5c9cd4d SW |
3759 | } else if (type == BTRFS_FILE_EXTENT_INLINE) { |
3760 | u64 skip = 0; | |
3761 | u64 trim = 0; | |
ed958762 | 3762 | |
c5c9cd4d SW |
3763 | if (off > key.offset) { |
3764 | skip = off - key.offset; | |
3765 | new_key.offset += skip; | |
3766 | } | |
d397712b | 3767 | |
aa42ffd9 LB |
3768 | if (key.offset + datal > off + len) |
3769 | trim = key.offset + datal - (off + len); | |
d397712b | 3770 | |
c5c9cd4d | 3771 | if (comp && (skip || trim)) { |
c5c9cd4d | 3772 | ret = -EINVAL; |
3a45bb20 | 3773 | btrfs_end_transaction(trans); |
c5c9cd4d SW |
3774 | goto out; |
3775 | } | |
3776 | size -= skip + trim; | |
3777 | datal -= skip + trim; | |
a22285a6 | 3778 | |
8039d87d FM |
3779 | ret = clone_copy_inline_extent(src, inode, |
3780 | trans, path, | |
3781 | &new_key, | |
3782 | drop_start, | |
3783 | datal, | |
3784 | skip, size, buf); | |
79787eaa | 3785 | if (ret) { |
3f9e3df8 | 3786 | if (ret != -EOPNOTSUPP) |
3a29bc09 | 3787 | btrfs_abort_transaction(trans, |
8039d87d | 3788 | ret); |
3a45bb20 | 3789 | btrfs_end_transaction(trans); |
79787eaa JM |
3790 | goto out; |
3791 | } | |
c5c9cd4d SW |
3792 | leaf = path->nodes[0]; |
3793 | slot = path->slots[0]; | |
f46b5a66 | 3794 | } |
c5c9cd4d | 3795 | |
7ffbb598 FM |
3796 | /* If we have an implicit hole (NO_HOLES feature). */ |
3797 | if (drop_start < new_key.offset) | |
3798 | clone_update_extent_map(inode, trans, | |
14f59796 | 3799 | NULL, drop_start, |
7ffbb598 FM |
3800 | new_key.offset - drop_start); |
3801 | ||
14f59796 | 3802 | clone_update_extent_map(inode, trans, path, 0, 0); |
7ffbb598 | 3803 | |
c5c9cd4d | 3804 | btrfs_mark_buffer_dirty(leaf); |
b3b4aa74 | 3805 | btrfs_release_path(path); |
c5c9cd4d | 3806 | |
62e2390e | 3807 | last_dest_end = ALIGN(new_key.offset + datal, |
0b246afa | 3808 | fs_info->sectorsize); |
f82a9901 FM |
3809 | ret = clone_finish_inode_update(trans, inode, |
3810 | last_dest_end, | |
1c919a5e MF |
3811 | destoff, olen, |
3812 | no_time_update); | |
f82a9901 | 3813 | if (ret) |
79787eaa | 3814 | goto out; |
2c463823 FM |
3815 | if (new_key.offset + datal >= destoff + len) |
3816 | break; | |
a22285a6 | 3817 | } |
b3b4aa74 | 3818 | btrfs_release_path(path); |
df858e76 | 3819 | key.offset = next_key_min_offset; |
69ae5e44 WX |
3820 | |
3821 | if (fatal_signal_pending(current)) { | |
3822 | ret = -EINTR; | |
3823 | goto out; | |
3824 | } | |
f46b5a66 | 3825 | } |
f46b5a66 | 3826 | ret = 0; |
32b7c687 | 3827 | |
f82a9901 FM |
3828 | if (last_dest_end < destoff + len) { |
3829 | /* | |
3830 | * We have an implicit hole (NO_HOLES feature is enabled) that | |
3831 | * fully or partially overlaps our cloning range at its end. | |
3832 | */ | |
3833 | btrfs_release_path(path); | |
3834 | ||
3835 | /* | |
3836 | * 1 - remove extent(s) | |
3837 | * 1 - inode update | |
3838 | */ | |
3839 | trans = btrfs_start_transaction(root, 2); | |
3840 | if (IS_ERR(trans)) { | |
3841 | ret = PTR_ERR(trans); | |
3842 | goto out; | |
3843 | } | |
3844 | ret = btrfs_drop_extents(trans, root, inode, | |
3845 | last_dest_end, destoff + len, 1); | |
3846 | if (ret) { | |
3847 | if (ret != -EOPNOTSUPP) | |
66642832 | 3848 | btrfs_abort_transaction(trans, ret); |
3a45bb20 | 3849 | btrfs_end_transaction(trans); |
f82a9901 FM |
3850 | goto out; |
3851 | } | |
14f59796 FM |
3852 | clone_update_extent_map(inode, trans, NULL, last_dest_end, |
3853 | destoff + len - last_dest_end); | |
f82a9901 | 3854 | ret = clone_finish_inode_update(trans, inode, destoff + len, |
1c919a5e | 3855 | destoff, olen, no_time_update); |
f82a9901 FM |
3856 | } |
3857 | ||
f46b5a66 | 3858 | out: |
32b7c687 | 3859 | btrfs_free_path(path); |
15351955 | 3860 | kvfree(buf); |
32b7c687 MF |
3861 | return ret; |
3862 | } | |
3863 | ||
3db11b2e ZB |
3864 | static noinline int btrfs_clone_files(struct file *file, struct file *file_src, |
3865 | u64 off, u64 olen, u64 destoff) | |
32b7c687 | 3866 | { |
54563d41 | 3867 | struct inode *inode = file_inode(file); |
3db11b2e | 3868 | struct inode *src = file_inode(file_src); |
0b246afa | 3869 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
32b7c687 | 3870 | struct btrfs_root *root = BTRFS_I(inode)->root; |
32b7c687 MF |
3871 | int ret; |
3872 | u64 len = olen; | |
0b246afa | 3873 | u64 bs = fs_info->sb->s_blocksize; |
3db11b2e | 3874 | int same_inode = src == inode; |
32b7c687 MF |
3875 | |
3876 | /* | |
3877 | * TODO: | |
3878 | * - split compressed inline extents. annoying: we need to | |
3879 | * decompress into destination's address_space (the file offset | |
3880 | * may change, so source mapping won't do), then recompress (or | |
3881 | * otherwise reinsert) a subrange. | |
00fdf13a LB |
3882 | * |
3883 | * - split destination inode's inline extents. The inline extents can | |
3884 | * be either compressed or non-compressed. | |
32b7c687 MF |
3885 | */ |
3886 | ||
32b7c687 MF |
3887 | if (btrfs_root_readonly(root)) |
3888 | return -EROFS; | |
3889 | ||
3db11b2e ZB |
3890 | if (file_src->f_path.mnt != file->f_path.mnt || |
3891 | src->i_sb != inode->i_sb) | |
3892 | return -EXDEV; | |
32b7c687 MF |
3893 | |
3894 | /* don't make the dst file partly checksummed */ | |
3895 | if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != | |
3896 | (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) | |
3db11b2e | 3897 | return -EINVAL; |
32b7c687 | 3898 | |
32b7c687 | 3899 | if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) |
3db11b2e | 3900 | return -EISDIR; |
32b7c687 MF |
3901 | |
3902 | if (!same_inode) { | |
293a8489 | 3903 | btrfs_double_inode_lock(src, inode); |
32b7c687 | 3904 | } else { |
5955102c | 3905 | inode_lock(src); |
32b7c687 MF |
3906 | } |
3907 | ||
3908 | /* determine range to clone */ | |
3909 | ret = -EINVAL; | |
3910 | if (off + len > src->i_size || off + len < off) | |
3911 | goto out_unlock; | |
3912 | if (len == 0) | |
3913 | olen = len = src->i_size - off; | |
3914 | /* if we extend to eof, continue to block boundary */ | |
3915 | if (off + len == src->i_size) | |
3916 | len = ALIGN(src->i_size, bs) - off; | |
3917 | ||
ccccf3d6 FM |
3918 | if (len == 0) { |
3919 | ret = 0; | |
3920 | goto out_unlock; | |
3921 | } | |
3922 | ||
32b7c687 MF |
3923 | /* verify the end result is block aligned */ |
3924 | if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || | |
3925 | !IS_ALIGNED(destoff, bs)) | |
3926 | goto out_unlock; | |
3927 | ||
3928 | /* verify if ranges are overlapped within the same file */ | |
3929 | if (same_inode) { | |
3930 | if (destoff + len > off && destoff < off + len) | |
3931 | goto out_unlock; | |
3932 | } | |
3933 | ||
3934 | if (destoff > inode->i_size) { | |
3935 | ret = btrfs_cont_expand(inode, inode->i_size, destoff); | |
3936 | if (ret) | |
3937 | goto out_unlock; | |
3938 | } | |
3939 | ||
c125b8bf FM |
3940 | /* |
3941 | * Lock the target range too. Right after we replace the file extent | |
3942 | * items in the fs tree (which now point to the cloned data), we might | |
3943 | * have a worker replace them with extent items relative to a write | |
3944 | * operation that was issued before this clone operation (i.e. confront | |
3945 | * with inode.c:btrfs_finish_ordered_io). | |
3946 | */ | |
3947 | if (same_inode) { | |
3948 | u64 lock_start = min_t(u64, off, destoff); | |
3949 | u64 lock_len = max_t(u64, off, destoff) + len - lock_start; | |
32b7c687 | 3950 | |
e0bd70c6 | 3951 | ret = lock_extent_range(src, lock_start, lock_len, true); |
c125b8bf | 3952 | } else { |
e0bd70c6 FM |
3953 | ret = btrfs_double_extent_lock(src, off, inode, destoff, len, |
3954 | true); | |
3955 | } | |
3956 | ASSERT(ret == 0); | |
3957 | if (WARN_ON(ret)) { | |
3958 | /* ranges in the io trees already unlocked */ | |
3959 | goto out_unlock; | |
c125b8bf | 3960 | } |
32b7c687 | 3961 | |
1c919a5e | 3962 | ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); |
32b7c687 | 3963 | |
c125b8bf FM |
3964 | if (same_inode) { |
3965 | u64 lock_start = min_t(u64, off, destoff); | |
3966 | u64 lock_end = max_t(u64, off, destoff) + len - 1; | |
3967 | ||
3968 | unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end); | |
3969 | } else { | |
293a8489 | 3970 | btrfs_double_extent_unlock(src, off, inode, destoff, len); |
c125b8bf FM |
3971 | } |
3972 | /* | |
3973 | * Truncate page cache pages so that future reads will see the cloned | |
3974 | * data immediately and not the previous data. | |
3975 | */ | |
65bfa658 | 3976 | truncate_inode_pages_range(&inode->i_data, |
09cbfeaf KS |
3977 | round_down(destoff, PAGE_SIZE), |
3978 | round_up(destoff + len, PAGE_SIZE) - 1); | |
f46b5a66 | 3979 | out_unlock: |
293a8489 MF |
3980 | if (!same_inode) |
3981 | btrfs_double_inode_unlock(src, inode); | |
3982 | else | |
5955102c | 3983 | inode_unlock(src); |
3db11b2e ZB |
3984 | return ret; |
3985 | } | |
3986 | ||
04b38d60 CH |
3987 | int btrfs_clone_file_range(struct file *src_file, loff_t off, |
3988 | struct file *dst_file, loff_t destoff, u64 len) | |
3db11b2e | 3989 | { |
04b38d60 | 3990 | return btrfs_clone_files(dst_file, src_file, off, len, destoff); |
c5c9cd4d SW |
3991 | } |
3992 | ||
f46b5a66 CH |
3993 | /* |
3994 | * there are many ways the trans_start and trans_end ioctls can lead | |
3995 | * to deadlocks. They should only be used by applications that | |
3996 | * basically own the machine, and have a very in depth understanding | |
3997 | * of all the possible deadlocks and enospc problems. | |
3998 | */ | |
b2950863 | 3999 | static long btrfs_ioctl_trans_start(struct file *file) |
f46b5a66 | 4000 | { |
496ad9aa | 4001 | struct inode *inode = file_inode(file); |
0b246afa | 4002 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
f46b5a66 CH |
4003 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4004 | struct btrfs_trans_handle *trans; | |
1ab86aed | 4005 | int ret; |
f46b5a66 | 4006 | |
1ab86aed | 4007 | ret = -EPERM; |
df5b5520 | 4008 | if (!capable(CAP_SYS_ADMIN)) |
1ab86aed | 4009 | goto out; |
df5b5520 | 4010 | |
1ab86aed SW |
4011 | ret = -EINPROGRESS; |
4012 | if (file->private_data) | |
f46b5a66 | 4013 | goto out; |
9ca9ee09 | 4014 | |
b83cc969 LZ |
4015 | ret = -EROFS; |
4016 | if (btrfs_root_readonly(root)) | |
4017 | goto out; | |
4018 | ||
a561be71 | 4019 | ret = mnt_want_write_file(file); |
c146afad YZ |
4020 | if (ret) |
4021 | goto out; | |
4022 | ||
0b246afa | 4023 | atomic_inc(&fs_info->open_ioctl_trans); |
9ca9ee09 | 4024 | |
1ab86aed | 4025 | ret = -ENOMEM; |
7a7eaa40 | 4026 | trans = btrfs_start_ioctl_transaction(root); |
abd30bb0 | 4027 | if (IS_ERR(trans)) |
1ab86aed SW |
4028 | goto out_drop; |
4029 | ||
4030 | file->private_data = trans; | |
4031 | return 0; | |
4032 | ||
4033 | out_drop: | |
0b246afa | 4034 | atomic_dec(&fs_info->open_ioctl_trans); |
2a79f17e | 4035 | mnt_drop_write_file(file); |
f46b5a66 | 4036 | out: |
f46b5a66 CH |
4037 | return ret; |
4038 | } | |
4039 | ||
6ef5ed0d JB |
4040 | static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) |
4041 | { | |
496ad9aa | 4042 | struct inode *inode = file_inode(file); |
0b246afa | 4043 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
6ef5ed0d JB |
4044 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4045 | struct btrfs_root *new_root; | |
4046 | struct btrfs_dir_item *di; | |
4047 | struct btrfs_trans_handle *trans; | |
4048 | struct btrfs_path *path; | |
4049 | struct btrfs_key location; | |
4050 | struct btrfs_disk_key disk_key; | |
6ef5ed0d JB |
4051 | u64 objectid = 0; |
4052 | u64 dir_id; | |
3c04ce01 | 4053 | int ret; |
6ef5ed0d JB |
4054 | |
4055 | if (!capable(CAP_SYS_ADMIN)) | |
4056 | return -EPERM; | |
4057 | ||
3c04ce01 MX |
4058 | ret = mnt_want_write_file(file); |
4059 | if (ret) | |
4060 | return ret; | |
4061 | ||
4062 | if (copy_from_user(&objectid, argp, sizeof(objectid))) { | |
4063 | ret = -EFAULT; | |
4064 | goto out; | |
4065 | } | |
6ef5ed0d JB |
4066 | |
4067 | if (!objectid) | |
1cecf579 | 4068 | objectid = BTRFS_FS_TREE_OBJECTID; |
6ef5ed0d JB |
4069 | |
4070 | location.objectid = objectid; | |
4071 | location.type = BTRFS_ROOT_ITEM_KEY; | |
4072 | location.offset = (u64)-1; | |
4073 | ||
0b246afa | 4074 | new_root = btrfs_read_fs_root_no_name(fs_info, &location); |
3c04ce01 MX |
4075 | if (IS_ERR(new_root)) { |
4076 | ret = PTR_ERR(new_root); | |
4077 | goto out; | |
4078 | } | |
6ef5ed0d | 4079 | |
6ef5ed0d | 4080 | path = btrfs_alloc_path(); |
3c04ce01 MX |
4081 | if (!path) { |
4082 | ret = -ENOMEM; | |
4083 | goto out; | |
4084 | } | |
6ef5ed0d JB |
4085 | path->leave_spinning = 1; |
4086 | ||
4087 | trans = btrfs_start_transaction(root, 1); | |
98d5dc13 | 4088 | if (IS_ERR(trans)) { |
6ef5ed0d | 4089 | btrfs_free_path(path); |
3c04ce01 MX |
4090 | ret = PTR_ERR(trans); |
4091 | goto out; | |
6ef5ed0d JB |
4092 | } |
4093 | ||
0b246afa JM |
4094 | dir_id = btrfs_super_root_dir(fs_info->super_copy); |
4095 | di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, | |
6ef5ed0d | 4096 | dir_id, "default", 7, 1); |
cf1e99a4 | 4097 | if (IS_ERR_OR_NULL(di)) { |
6ef5ed0d | 4098 | btrfs_free_path(path); |
3a45bb20 | 4099 | btrfs_end_transaction(trans); |
0b246afa | 4100 | btrfs_err(fs_info, |
5d163e0e | 4101 | "Umm, you don't have the default diritem, this isn't going to work"); |
3c04ce01 MX |
4102 | ret = -ENOENT; |
4103 | goto out; | |
6ef5ed0d JB |
4104 | } |
4105 | ||
4106 | btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); | |
4107 | btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); | |
4108 | btrfs_mark_buffer_dirty(path->nodes[0]); | |
4109 | btrfs_free_path(path); | |
4110 | ||
0b246afa | 4111 | btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); |
3a45bb20 | 4112 | btrfs_end_transaction(trans); |
3c04ce01 MX |
4113 | out: |
4114 | mnt_drop_write_file(file); | |
4115 | return ret; | |
6ef5ed0d JB |
4116 | } |
4117 | ||
5af3e8cc SB |
4118 | void btrfs_get_block_group_info(struct list_head *groups_list, |
4119 | struct btrfs_ioctl_space_info *space) | |
bf5fc093 JB |
4120 | { |
4121 | struct btrfs_block_group_cache *block_group; | |
4122 | ||
4123 | space->total_bytes = 0; | |
4124 | space->used_bytes = 0; | |
4125 | space->flags = 0; | |
4126 | list_for_each_entry(block_group, groups_list, list) { | |
4127 | space->flags = block_group->flags; | |
4128 | space->total_bytes += block_group->key.offset; | |
4129 | space->used_bytes += | |
4130 | btrfs_block_group_used(&block_group->item); | |
4131 | } | |
4132 | } | |
4133 | ||
2ff7e61e JM |
4134 | static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, |
4135 | void __user *arg) | |
1406e432 JB |
4136 | { |
4137 | struct btrfs_ioctl_space_args space_args; | |
4138 | struct btrfs_ioctl_space_info space; | |
4139 | struct btrfs_ioctl_space_info *dest; | |
7fde62bf | 4140 | struct btrfs_ioctl_space_info *dest_orig; |
13f2696f | 4141 | struct btrfs_ioctl_space_info __user *user_dest; |
1406e432 | 4142 | struct btrfs_space_info *info; |
bf5fc093 JB |
4143 | u64 types[] = {BTRFS_BLOCK_GROUP_DATA, |
4144 | BTRFS_BLOCK_GROUP_SYSTEM, | |
4145 | BTRFS_BLOCK_GROUP_METADATA, | |
4146 | BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; | |
4147 | int num_types = 4; | |
7fde62bf | 4148 | int alloc_size; |
1406e432 | 4149 | int ret = 0; |
51788b1b | 4150 | u64 slot_count = 0; |
bf5fc093 | 4151 | int i, c; |
1406e432 JB |
4152 | |
4153 | if (copy_from_user(&space_args, | |
4154 | (struct btrfs_ioctl_space_args __user *)arg, | |
4155 | sizeof(space_args))) | |
4156 | return -EFAULT; | |
4157 | ||
bf5fc093 JB |
4158 | for (i = 0; i < num_types; i++) { |
4159 | struct btrfs_space_info *tmp; | |
4160 | ||
4161 | info = NULL; | |
4162 | rcu_read_lock(); | |
0b246afa | 4163 | list_for_each_entry_rcu(tmp, &fs_info->space_info, |
bf5fc093 JB |
4164 | list) { |
4165 | if (tmp->flags == types[i]) { | |
4166 | info = tmp; | |
4167 | break; | |
4168 | } | |
4169 | } | |
4170 | rcu_read_unlock(); | |
4171 | ||
4172 | if (!info) | |
4173 | continue; | |
4174 | ||
4175 | down_read(&info->groups_sem); | |
4176 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
4177 | if (!list_empty(&info->block_groups[c])) | |
4178 | slot_count++; | |
4179 | } | |
4180 | up_read(&info->groups_sem); | |
4181 | } | |
7fde62bf | 4182 | |
36523e95 DS |
4183 | /* |
4184 | * Global block reserve, exported as a space_info | |
4185 | */ | |
4186 | slot_count++; | |
4187 | ||
7fde62bf CM |
4188 | /* space_slots == 0 means they are asking for a count */ |
4189 | if (space_args.space_slots == 0) { | |
4190 | space_args.total_spaces = slot_count; | |
4191 | goto out; | |
4192 | } | |
bf5fc093 | 4193 | |
51788b1b | 4194 | slot_count = min_t(u64, space_args.space_slots, slot_count); |
bf5fc093 | 4195 | |
7fde62bf | 4196 | alloc_size = sizeof(*dest) * slot_count; |
bf5fc093 | 4197 | |
7fde62bf CM |
4198 | /* we generally have at most 6 or so space infos, one for each raid |
4199 | * level. So, a whole page should be more than enough for everyone | |
4200 | */ | |
09cbfeaf | 4201 | if (alloc_size > PAGE_SIZE) |
7fde62bf CM |
4202 | return -ENOMEM; |
4203 | ||
1406e432 | 4204 | space_args.total_spaces = 0; |
8d2db785 | 4205 | dest = kmalloc(alloc_size, GFP_KERNEL); |
7fde62bf CM |
4206 | if (!dest) |
4207 | return -ENOMEM; | |
4208 | dest_orig = dest; | |
1406e432 | 4209 | |
7fde62bf | 4210 | /* now we have a buffer to copy into */ |
bf5fc093 JB |
4211 | for (i = 0; i < num_types; i++) { |
4212 | struct btrfs_space_info *tmp; | |
4213 | ||
51788b1b DR |
4214 | if (!slot_count) |
4215 | break; | |
4216 | ||
bf5fc093 JB |
4217 | info = NULL; |
4218 | rcu_read_lock(); | |
0b246afa | 4219 | list_for_each_entry_rcu(tmp, &fs_info->space_info, |
bf5fc093 JB |
4220 | list) { |
4221 | if (tmp->flags == types[i]) { | |
4222 | info = tmp; | |
4223 | break; | |
4224 | } | |
4225 | } | |
4226 | rcu_read_unlock(); | |
7fde62bf | 4227 | |
bf5fc093 JB |
4228 | if (!info) |
4229 | continue; | |
4230 | down_read(&info->groups_sem); | |
4231 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
4232 | if (!list_empty(&info->block_groups[c])) { | |
5af3e8cc SB |
4233 | btrfs_get_block_group_info( |
4234 | &info->block_groups[c], &space); | |
bf5fc093 JB |
4235 | memcpy(dest, &space, sizeof(space)); |
4236 | dest++; | |
4237 | space_args.total_spaces++; | |
51788b1b | 4238 | slot_count--; |
bf5fc093 | 4239 | } |
51788b1b DR |
4240 | if (!slot_count) |
4241 | break; | |
bf5fc093 JB |
4242 | } |
4243 | up_read(&info->groups_sem); | |
1406e432 | 4244 | } |
1406e432 | 4245 | |
36523e95 DS |
4246 | /* |
4247 | * Add global block reserve | |
4248 | */ | |
4249 | if (slot_count) { | |
0b246afa | 4250 | struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; |
36523e95 DS |
4251 | |
4252 | spin_lock(&block_rsv->lock); | |
4253 | space.total_bytes = block_rsv->size; | |
4254 | space.used_bytes = block_rsv->size - block_rsv->reserved; | |
4255 | spin_unlock(&block_rsv->lock); | |
4256 | space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; | |
4257 | memcpy(dest, &space, sizeof(space)); | |
4258 | space_args.total_spaces++; | |
4259 | } | |
4260 | ||
2eec6c81 | 4261 | user_dest = (struct btrfs_ioctl_space_info __user *) |
7fde62bf CM |
4262 | (arg + sizeof(struct btrfs_ioctl_space_args)); |
4263 | ||
4264 | if (copy_to_user(user_dest, dest_orig, alloc_size)) | |
4265 | ret = -EFAULT; | |
4266 | ||
4267 | kfree(dest_orig); | |
4268 | out: | |
4269 | if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) | |
1406e432 JB |
4270 | ret = -EFAULT; |
4271 | ||
4272 | return ret; | |
4273 | } | |
4274 | ||
f46b5a66 CH |
4275 | /* |
4276 | * there are many ways the trans_start and trans_end ioctls can lead | |
4277 | * to deadlocks. They should only be used by applications that | |
4278 | * basically own the machine, and have a very in depth understanding | |
4279 | * of all the possible deadlocks and enospc problems. | |
4280 | */ | |
4281 | long btrfs_ioctl_trans_end(struct file *file) | |
4282 | { | |
496ad9aa | 4283 | struct inode *inode = file_inode(file); |
f46b5a66 CH |
4284 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4285 | struct btrfs_trans_handle *trans; | |
f46b5a66 | 4286 | |
f46b5a66 | 4287 | trans = file->private_data; |
1ab86aed SW |
4288 | if (!trans) |
4289 | return -EINVAL; | |
b214107e | 4290 | file->private_data = NULL; |
9ca9ee09 | 4291 | |
3a45bb20 | 4292 | btrfs_end_transaction(trans); |
1ab86aed | 4293 | |
a4abeea4 | 4294 | atomic_dec(&root->fs_info->open_ioctl_trans); |
9ca9ee09 | 4295 | |
2a79f17e | 4296 | mnt_drop_write_file(file); |
1ab86aed | 4297 | return 0; |
f46b5a66 CH |
4298 | } |
4299 | ||
9a8c28be MX |
4300 | static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, |
4301 | void __user *argp) | |
46204592 | 4302 | { |
46204592 SW |
4303 | struct btrfs_trans_handle *trans; |
4304 | u64 transid; | |
db5b493a | 4305 | int ret; |
46204592 | 4306 | |
d4edf39b | 4307 | trans = btrfs_attach_transaction_barrier(root); |
ff7c1d33 MX |
4308 | if (IS_ERR(trans)) { |
4309 | if (PTR_ERR(trans) != -ENOENT) | |
4310 | return PTR_ERR(trans); | |
4311 | ||
4312 | /* No running transaction, don't bother */ | |
4313 | transid = root->fs_info->last_trans_committed; | |
4314 | goto out; | |
4315 | } | |
46204592 | 4316 | transid = trans->transid; |
3a45bb20 | 4317 | ret = btrfs_commit_transaction_async(trans, 0); |
8b2b2d3c | 4318 | if (ret) { |
3a45bb20 | 4319 | btrfs_end_transaction(trans); |
db5b493a | 4320 | return ret; |
8b2b2d3c | 4321 | } |
ff7c1d33 | 4322 | out: |
46204592 SW |
4323 | if (argp) |
4324 | if (copy_to_user(argp, &transid, sizeof(transid))) | |
4325 | return -EFAULT; | |
4326 | return 0; | |
4327 | } | |
4328 | ||
2ff7e61e | 4329 | static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, |
9a8c28be | 4330 | void __user *argp) |
46204592 | 4331 | { |
46204592 SW |
4332 | u64 transid; |
4333 | ||
4334 | if (argp) { | |
4335 | if (copy_from_user(&transid, argp, sizeof(transid))) | |
4336 | return -EFAULT; | |
4337 | } else { | |
4338 | transid = 0; /* current trans */ | |
4339 | } | |
2ff7e61e | 4340 | return btrfs_wait_for_commit(fs_info, transid); |
46204592 SW |
4341 | } |
4342 | ||
b8e95489 | 4343 | static long btrfs_ioctl_scrub(struct file *file, void __user *arg) |
475f6387 | 4344 | { |
0b246afa | 4345 | struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); |
475f6387 | 4346 | struct btrfs_ioctl_scrub_args *sa; |
b8e95489 | 4347 | int ret; |
475f6387 JS |
4348 | |
4349 | if (!capable(CAP_SYS_ADMIN)) | |
4350 | return -EPERM; | |
4351 | ||
4352 | sa = memdup_user(arg, sizeof(*sa)); | |
4353 | if (IS_ERR(sa)) | |
4354 | return PTR_ERR(sa); | |
4355 | ||
b8e95489 MX |
4356 | if (!(sa->flags & BTRFS_SCRUB_READONLY)) { |
4357 | ret = mnt_want_write_file(file); | |
4358 | if (ret) | |
4359 | goto out; | |
4360 | } | |
4361 | ||
0b246afa | 4362 | ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, |
63a212ab SB |
4363 | &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, |
4364 | 0); | |
475f6387 JS |
4365 | |
4366 | if (copy_to_user(arg, sa, sizeof(*sa))) | |
4367 | ret = -EFAULT; | |
4368 | ||
b8e95489 MX |
4369 | if (!(sa->flags & BTRFS_SCRUB_READONLY)) |
4370 | mnt_drop_write_file(file); | |
4371 | out: | |
475f6387 JS |
4372 | kfree(sa); |
4373 | return ret; | |
4374 | } | |
4375 | ||
2ff7e61e | 4376 | static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) |
475f6387 JS |
4377 | { |
4378 | if (!capable(CAP_SYS_ADMIN)) | |
4379 | return -EPERM; | |
4380 | ||
2ff7e61e | 4381 | return btrfs_scrub_cancel(fs_info); |
475f6387 JS |
4382 | } |
4383 | ||
2ff7e61e | 4384 | static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, |
475f6387 JS |
4385 | void __user *arg) |
4386 | { | |
4387 | struct btrfs_ioctl_scrub_args *sa; | |
4388 | int ret; | |
4389 | ||
4390 | if (!capable(CAP_SYS_ADMIN)) | |
4391 | return -EPERM; | |
4392 | ||
4393 | sa = memdup_user(arg, sizeof(*sa)); | |
4394 | if (IS_ERR(sa)) | |
4395 | return PTR_ERR(sa); | |
4396 | ||
2ff7e61e | 4397 | ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); |
475f6387 JS |
4398 | |
4399 | if (copy_to_user(arg, sa, sizeof(*sa))) | |
4400 | ret = -EFAULT; | |
4401 | ||
4402 | kfree(sa); | |
4403 | return ret; | |
4404 | } | |
4405 | ||
2ff7e61e | 4406 | static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, |
b27f7c0c | 4407 | void __user *arg) |
c11d2c23 SB |
4408 | { |
4409 | struct btrfs_ioctl_get_dev_stats *sa; | |
4410 | int ret; | |
4411 | ||
c11d2c23 SB |
4412 | sa = memdup_user(arg, sizeof(*sa)); |
4413 | if (IS_ERR(sa)) | |
4414 | return PTR_ERR(sa); | |
4415 | ||
b27f7c0c DS |
4416 | if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { |
4417 | kfree(sa); | |
4418 | return -EPERM; | |
4419 | } | |
4420 | ||
2ff7e61e | 4421 | ret = btrfs_get_dev_stats(fs_info, sa); |
c11d2c23 SB |
4422 | |
4423 | if (copy_to_user(arg, sa, sizeof(*sa))) | |
4424 | ret = -EFAULT; | |
4425 | ||
4426 | kfree(sa); | |
4427 | return ret; | |
4428 | } | |
4429 | ||
2ff7e61e JM |
4430 | static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, |
4431 | void __user *arg) | |
3f6bcfbd SB |
4432 | { |
4433 | struct btrfs_ioctl_dev_replace_args *p; | |
4434 | int ret; | |
4435 | ||
4436 | if (!capable(CAP_SYS_ADMIN)) | |
4437 | return -EPERM; | |
4438 | ||
4439 | p = memdup_user(arg, sizeof(*p)); | |
4440 | if (IS_ERR(p)) | |
4441 | return PTR_ERR(p); | |
4442 | ||
4443 | switch (p->cmd) { | |
4444 | case BTRFS_IOCTL_DEV_REPLACE_CMD_START: | |
0b246afa | 4445 | if (fs_info->sb->s_flags & MS_RDONLY) { |
adfa97cb ID |
4446 | ret = -EROFS; |
4447 | goto out; | |
4448 | } | |
3f6bcfbd | 4449 | if (atomic_xchg( |
0b246afa | 4450 | &fs_info->mutually_exclusive_operation_running, 1)) { |
e57138b3 | 4451 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
3f6bcfbd | 4452 | } else { |
2ff7e61e | 4453 | ret = btrfs_dev_replace_by_ioctl(fs_info, p); |
3f6bcfbd | 4454 | atomic_set( |
0b246afa | 4455 | &fs_info->mutually_exclusive_operation_running, 0); |
3f6bcfbd SB |
4456 | } |
4457 | break; | |
4458 | case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: | |
0b246afa | 4459 | btrfs_dev_replace_status(fs_info, p); |
3f6bcfbd SB |
4460 | ret = 0; |
4461 | break; | |
4462 | case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: | |
0b246afa | 4463 | ret = btrfs_dev_replace_cancel(fs_info, p); |
3f6bcfbd SB |
4464 | break; |
4465 | default: | |
4466 | ret = -EINVAL; | |
4467 | break; | |
4468 | } | |
4469 | ||
4470 | if (copy_to_user(arg, p, sizeof(*p))) | |
4471 | ret = -EFAULT; | |
adfa97cb | 4472 | out: |
3f6bcfbd SB |
4473 | kfree(p); |
4474 | return ret; | |
4475 | } | |
4476 | ||
d7728c96 JS |
4477 | static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) |
4478 | { | |
4479 | int ret = 0; | |
4480 | int i; | |
740c3d22 | 4481 | u64 rel_ptr; |
d7728c96 | 4482 | int size; |
806468f8 | 4483 | struct btrfs_ioctl_ino_path_args *ipa = NULL; |
d7728c96 JS |
4484 | struct inode_fs_paths *ipath = NULL; |
4485 | struct btrfs_path *path; | |
4486 | ||
82b22ac8 | 4487 | if (!capable(CAP_DAC_READ_SEARCH)) |
d7728c96 JS |
4488 | return -EPERM; |
4489 | ||
4490 | path = btrfs_alloc_path(); | |
4491 | if (!path) { | |
4492 | ret = -ENOMEM; | |
4493 | goto out; | |
4494 | } | |
4495 | ||
4496 | ipa = memdup_user(arg, sizeof(*ipa)); | |
4497 | if (IS_ERR(ipa)) { | |
4498 | ret = PTR_ERR(ipa); | |
4499 | ipa = NULL; | |
4500 | goto out; | |
4501 | } | |
4502 | ||
4503 | size = min_t(u32, ipa->size, 4096); | |
4504 | ipath = init_ipath(size, root, path); | |
4505 | if (IS_ERR(ipath)) { | |
4506 | ret = PTR_ERR(ipath); | |
4507 | ipath = NULL; | |
4508 | goto out; | |
4509 | } | |
4510 | ||
4511 | ret = paths_from_inode(ipa->inum, ipath); | |
4512 | if (ret < 0) | |
4513 | goto out; | |
4514 | ||
4515 | for (i = 0; i < ipath->fspath->elem_cnt; ++i) { | |
745c4d8e JM |
4516 | rel_ptr = ipath->fspath->val[i] - |
4517 | (u64)(unsigned long)ipath->fspath->val; | |
740c3d22 | 4518 | ipath->fspath->val[i] = rel_ptr; |
d7728c96 JS |
4519 | } |
4520 | ||
745c4d8e JM |
4521 | ret = copy_to_user((void *)(unsigned long)ipa->fspath, |
4522 | (void *)(unsigned long)ipath->fspath, size); | |
d7728c96 JS |
4523 | if (ret) { |
4524 | ret = -EFAULT; | |
4525 | goto out; | |
4526 | } | |
4527 | ||
4528 | out: | |
4529 | btrfs_free_path(path); | |
4530 | free_ipath(ipath); | |
4531 | kfree(ipa); | |
4532 | ||
4533 | return ret; | |
4534 | } | |
4535 | ||
4536 | static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) | |
4537 | { | |
4538 | struct btrfs_data_container *inodes = ctx; | |
4539 | const size_t c = 3 * sizeof(u64); | |
4540 | ||
4541 | if (inodes->bytes_left >= c) { | |
4542 | inodes->bytes_left -= c; | |
4543 | inodes->val[inodes->elem_cnt] = inum; | |
4544 | inodes->val[inodes->elem_cnt + 1] = offset; | |
4545 | inodes->val[inodes->elem_cnt + 2] = root; | |
4546 | inodes->elem_cnt += 3; | |
4547 | } else { | |
4548 | inodes->bytes_missing += c - inodes->bytes_left; | |
4549 | inodes->bytes_left = 0; | |
4550 | inodes->elem_missed += 3; | |
4551 | } | |
4552 | ||
4553 | return 0; | |
4554 | } | |
4555 | ||
2ff7e61e | 4556 | static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, |
d7728c96 JS |
4557 | void __user *arg) |
4558 | { | |
4559 | int ret = 0; | |
4560 | int size; | |
d7728c96 JS |
4561 | struct btrfs_ioctl_logical_ino_args *loi; |
4562 | struct btrfs_data_container *inodes = NULL; | |
4563 | struct btrfs_path *path = NULL; | |
d7728c96 JS |
4564 | |
4565 | if (!capable(CAP_SYS_ADMIN)) | |
4566 | return -EPERM; | |
4567 | ||
4568 | loi = memdup_user(arg, sizeof(*loi)); | |
7b9ea627 SV |
4569 | if (IS_ERR(loi)) |
4570 | return PTR_ERR(loi); | |
d7728c96 JS |
4571 | |
4572 | path = btrfs_alloc_path(); | |
4573 | if (!path) { | |
4574 | ret = -ENOMEM; | |
4575 | goto out; | |
4576 | } | |
4577 | ||
ee22184b | 4578 | size = min_t(u32, loi->size, SZ_64K); |
d7728c96 JS |
4579 | inodes = init_data_container(size); |
4580 | if (IS_ERR(inodes)) { | |
4581 | ret = PTR_ERR(inodes); | |
4582 | inodes = NULL; | |
4583 | goto out; | |
4584 | } | |
4585 | ||
2ff7e61e | 4586 | ret = iterate_inodes_from_logical(loi->logical, fs_info, path, |
df031f07 LB |
4587 | build_ino_list, inodes); |
4588 | if (ret == -EINVAL) | |
d7728c96 JS |
4589 | ret = -ENOENT; |
4590 | if (ret < 0) | |
4591 | goto out; | |
4592 | ||
745c4d8e JM |
4593 | ret = copy_to_user((void *)(unsigned long)loi->inodes, |
4594 | (void *)(unsigned long)inodes, size); | |
d7728c96 JS |
4595 | if (ret) |
4596 | ret = -EFAULT; | |
4597 | ||
4598 | out: | |
4599 | btrfs_free_path(path); | |
425d17a2 | 4600 | vfree(inodes); |
d7728c96 JS |
4601 | kfree(loi); |
4602 | ||
4603 | return ret; | |
4604 | } | |
4605 | ||
19a39dce | 4606 | void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock, |
c9e9f97b ID |
4607 | struct btrfs_ioctl_balance_args *bargs) |
4608 | { | |
4609 | struct btrfs_balance_control *bctl = fs_info->balance_ctl; | |
4610 | ||
4611 | bargs->flags = bctl->flags; | |
4612 | ||
837d5b6e ID |
4613 | if (atomic_read(&fs_info->balance_running)) |
4614 | bargs->state |= BTRFS_BALANCE_STATE_RUNNING; | |
4615 | if (atomic_read(&fs_info->balance_pause_req)) | |
4616 | bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; | |
a7e99c69 ID |
4617 | if (atomic_read(&fs_info->balance_cancel_req)) |
4618 | bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; | |
837d5b6e | 4619 | |
c9e9f97b ID |
4620 | memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); |
4621 | memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); | |
4622 | memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); | |
19a39dce ID |
4623 | |
4624 | if (lock) { | |
4625 | spin_lock(&fs_info->balance_lock); | |
4626 | memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); | |
4627 | spin_unlock(&fs_info->balance_lock); | |
4628 | } else { | |
4629 | memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); | |
4630 | } | |
c9e9f97b ID |
4631 | } |
4632 | ||
9ba1f6e4 | 4633 | static long btrfs_ioctl_balance(struct file *file, void __user *arg) |
c9e9f97b | 4634 | { |
496ad9aa | 4635 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
c9e9f97b ID |
4636 | struct btrfs_fs_info *fs_info = root->fs_info; |
4637 | struct btrfs_ioctl_balance_args *bargs; | |
4638 | struct btrfs_balance_control *bctl; | |
ed0fb78f | 4639 | bool need_unlock; /* for mut. excl. ops lock */ |
c9e9f97b ID |
4640 | int ret; |
4641 | ||
4642 | if (!capable(CAP_SYS_ADMIN)) | |
4643 | return -EPERM; | |
4644 | ||
e54bfa31 | 4645 | ret = mnt_want_write_file(file); |
9ba1f6e4 LB |
4646 | if (ret) |
4647 | return ret; | |
4648 | ||
ed0fb78f ID |
4649 | again: |
4650 | if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { | |
4651 | mutex_lock(&fs_info->volume_mutex); | |
4652 | mutex_lock(&fs_info->balance_mutex); | |
4653 | need_unlock = true; | |
4654 | goto locked; | |
4655 | } | |
4656 | ||
4657 | /* | |
01327610 | 4658 | * mut. excl. ops lock is locked. Three possibilities: |
ed0fb78f ID |
4659 | * (1) some other op is running |
4660 | * (2) balance is running | |
4661 | * (3) balance is paused -- special case (think resume) | |
4662 | */ | |
c9e9f97b | 4663 | mutex_lock(&fs_info->balance_mutex); |
ed0fb78f ID |
4664 | if (fs_info->balance_ctl) { |
4665 | /* this is either (2) or (3) */ | |
4666 | if (!atomic_read(&fs_info->balance_running)) { | |
4667 | mutex_unlock(&fs_info->balance_mutex); | |
4668 | if (!mutex_trylock(&fs_info->volume_mutex)) | |
4669 | goto again; | |
4670 | mutex_lock(&fs_info->balance_mutex); | |
4671 | ||
4672 | if (fs_info->balance_ctl && | |
4673 | !atomic_read(&fs_info->balance_running)) { | |
4674 | /* this is (3) */ | |
4675 | need_unlock = false; | |
4676 | goto locked; | |
4677 | } | |
4678 | ||
4679 | mutex_unlock(&fs_info->balance_mutex); | |
4680 | mutex_unlock(&fs_info->volume_mutex); | |
4681 | goto again; | |
4682 | } else { | |
4683 | /* this is (2) */ | |
4684 | mutex_unlock(&fs_info->balance_mutex); | |
4685 | ret = -EINPROGRESS; | |
4686 | goto out; | |
4687 | } | |
4688 | } else { | |
4689 | /* this is (1) */ | |
4690 | mutex_unlock(&fs_info->balance_mutex); | |
e57138b3 | 4691 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
ed0fb78f ID |
4692 | goto out; |
4693 | } | |
4694 | ||
4695 | locked: | |
4696 | BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running)); | |
c9e9f97b ID |
4697 | |
4698 | if (arg) { | |
4699 | bargs = memdup_user(arg, sizeof(*bargs)); | |
4700 | if (IS_ERR(bargs)) { | |
4701 | ret = PTR_ERR(bargs); | |
ed0fb78f | 4702 | goto out_unlock; |
c9e9f97b | 4703 | } |
de322263 ID |
4704 | |
4705 | if (bargs->flags & BTRFS_BALANCE_RESUME) { | |
4706 | if (!fs_info->balance_ctl) { | |
4707 | ret = -ENOTCONN; | |
4708 | goto out_bargs; | |
4709 | } | |
4710 | ||
4711 | bctl = fs_info->balance_ctl; | |
4712 | spin_lock(&fs_info->balance_lock); | |
4713 | bctl->flags |= BTRFS_BALANCE_RESUME; | |
4714 | spin_unlock(&fs_info->balance_lock); | |
4715 | ||
4716 | goto do_balance; | |
4717 | } | |
c9e9f97b ID |
4718 | } else { |
4719 | bargs = NULL; | |
4720 | } | |
4721 | ||
ed0fb78f | 4722 | if (fs_info->balance_ctl) { |
837d5b6e ID |
4723 | ret = -EINPROGRESS; |
4724 | goto out_bargs; | |
4725 | } | |
4726 | ||
8d2db785 | 4727 | bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); |
c9e9f97b ID |
4728 | if (!bctl) { |
4729 | ret = -ENOMEM; | |
4730 | goto out_bargs; | |
4731 | } | |
4732 | ||
4733 | bctl->fs_info = fs_info; | |
4734 | if (arg) { | |
4735 | memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); | |
4736 | memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); | |
4737 | memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); | |
4738 | ||
4739 | bctl->flags = bargs->flags; | |
f43ffb60 ID |
4740 | } else { |
4741 | /* balance everything - no filters */ | |
4742 | bctl->flags |= BTRFS_BALANCE_TYPE_MASK; | |
c9e9f97b ID |
4743 | } |
4744 | ||
8eb93459 DS |
4745 | if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { |
4746 | ret = -EINVAL; | |
0f89abf5 | 4747 | goto out_bctl; |
8eb93459 DS |
4748 | } |
4749 | ||
de322263 | 4750 | do_balance: |
c9e9f97b | 4751 | /* |
ed0fb78f ID |
4752 | * Ownership of bctl and mutually_exclusive_operation_running |
4753 | * goes to to btrfs_balance. bctl is freed in __cancel_balance, | |
4754 | * or, if restriper was paused all the way until unmount, in | |
4755 | * free_fs_info. mutually_exclusive_operation_running is | |
4756 | * cleared in __cancel_balance. | |
c9e9f97b | 4757 | */ |
ed0fb78f ID |
4758 | need_unlock = false; |
4759 | ||
4760 | ret = btrfs_balance(bctl, bargs); | |
0f89abf5 | 4761 | bctl = NULL; |
ed0fb78f | 4762 | |
c9e9f97b ID |
4763 | if (arg) { |
4764 | if (copy_to_user(arg, bargs, sizeof(*bargs))) | |
4765 | ret = -EFAULT; | |
4766 | } | |
4767 | ||
0f89abf5 CE |
4768 | out_bctl: |
4769 | kfree(bctl); | |
c9e9f97b ID |
4770 | out_bargs: |
4771 | kfree(bargs); | |
ed0fb78f | 4772 | out_unlock: |
c9e9f97b ID |
4773 | mutex_unlock(&fs_info->balance_mutex); |
4774 | mutex_unlock(&fs_info->volume_mutex); | |
ed0fb78f ID |
4775 | if (need_unlock) |
4776 | atomic_set(&fs_info->mutually_exclusive_operation_running, 0); | |
4777 | out: | |
e54bfa31 | 4778 | mnt_drop_write_file(file); |
c9e9f97b ID |
4779 | return ret; |
4780 | } | |
4781 | ||
2ff7e61e | 4782 | static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) |
837d5b6e ID |
4783 | { |
4784 | if (!capable(CAP_SYS_ADMIN)) | |
4785 | return -EPERM; | |
4786 | ||
4787 | switch (cmd) { | |
4788 | case BTRFS_BALANCE_CTL_PAUSE: | |
0b246afa | 4789 | return btrfs_pause_balance(fs_info); |
a7e99c69 | 4790 | case BTRFS_BALANCE_CTL_CANCEL: |
0b246afa | 4791 | return btrfs_cancel_balance(fs_info); |
837d5b6e ID |
4792 | } |
4793 | ||
4794 | return -EINVAL; | |
4795 | } | |
4796 | ||
2ff7e61e | 4797 | static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, |
19a39dce ID |
4798 | void __user *arg) |
4799 | { | |
19a39dce ID |
4800 | struct btrfs_ioctl_balance_args *bargs; |
4801 | int ret = 0; | |
4802 | ||
4803 | if (!capable(CAP_SYS_ADMIN)) | |
4804 | return -EPERM; | |
4805 | ||
4806 | mutex_lock(&fs_info->balance_mutex); | |
4807 | if (!fs_info->balance_ctl) { | |
4808 | ret = -ENOTCONN; | |
4809 | goto out; | |
4810 | } | |
4811 | ||
8d2db785 | 4812 | bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); |
19a39dce ID |
4813 | if (!bargs) { |
4814 | ret = -ENOMEM; | |
4815 | goto out; | |
4816 | } | |
4817 | ||
4818 | update_ioctl_balance_args(fs_info, 1, bargs); | |
4819 | ||
4820 | if (copy_to_user(arg, bargs, sizeof(*bargs))) | |
4821 | ret = -EFAULT; | |
4822 | ||
4823 | kfree(bargs); | |
4824 | out: | |
4825 | mutex_unlock(&fs_info->balance_mutex); | |
4826 | return ret; | |
4827 | } | |
4828 | ||
905b0dda | 4829 | static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) |
5d13a37b | 4830 | { |
0b246afa JM |
4831 | struct inode *inode = file_inode(file); |
4832 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
5d13a37b AJ |
4833 | struct btrfs_ioctl_quota_ctl_args *sa; |
4834 | struct btrfs_trans_handle *trans = NULL; | |
4835 | int ret; | |
4836 | int err; | |
4837 | ||
4838 | if (!capable(CAP_SYS_ADMIN)) | |
4839 | return -EPERM; | |
4840 | ||
905b0dda MX |
4841 | ret = mnt_want_write_file(file); |
4842 | if (ret) | |
4843 | return ret; | |
5d13a37b AJ |
4844 | |
4845 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4846 | if (IS_ERR(sa)) { |
4847 | ret = PTR_ERR(sa); | |
4848 | goto drop_write; | |
4849 | } | |
5d13a37b | 4850 | |
0b246afa JM |
4851 | down_write(&fs_info->subvol_sem); |
4852 | trans = btrfs_start_transaction(fs_info->tree_root, 2); | |
2f232036 JS |
4853 | if (IS_ERR(trans)) { |
4854 | ret = PTR_ERR(trans); | |
4855 | goto out; | |
5d13a37b AJ |
4856 | } |
4857 | ||
4858 | switch (sa->cmd) { | |
4859 | case BTRFS_QUOTA_CTL_ENABLE: | |
0b246afa | 4860 | ret = btrfs_quota_enable(trans, fs_info); |
5d13a37b AJ |
4861 | break; |
4862 | case BTRFS_QUOTA_CTL_DISABLE: | |
0b246afa | 4863 | ret = btrfs_quota_disable(trans, fs_info); |
5d13a37b | 4864 | break; |
5d13a37b AJ |
4865 | default: |
4866 | ret = -EINVAL; | |
4867 | break; | |
4868 | } | |
4869 | ||
3a45bb20 | 4870 | err = btrfs_commit_transaction(trans); |
2f232036 JS |
4871 | if (err && !ret) |
4872 | ret = err; | |
5d13a37b AJ |
4873 | out: |
4874 | kfree(sa); | |
0b246afa | 4875 | up_write(&fs_info->subvol_sem); |
905b0dda MX |
4876 | drop_write: |
4877 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4878 | return ret; |
4879 | } | |
4880 | ||
905b0dda | 4881 | static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) |
5d13a37b | 4882 | { |
0b246afa JM |
4883 | struct inode *inode = file_inode(file); |
4884 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
4885 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
5d13a37b AJ |
4886 | struct btrfs_ioctl_qgroup_assign_args *sa; |
4887 | struct btrfs_trans_handle *trans; | |
4888 | int ret; | |
4889 | int err; | |
4890 | ||
4891 | if (!capable(CAP_SYS_ADMIN)) | |
4892 | return -EPERM; | |
4893 | ||
905b0dda MX |
4894 | ret = mnt_want_write_file(file); |
4895 | if (ret) | |
4896 | return ret; | |
5d13a37b AJ |
4897 | |
4898 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4899 | if (IS_ERR(sa)) { |
4900 | ret = PTR_ERR(sa); | |
4901 | goto drop_write; | |
4902 | } | |
5d13a37b AJ |
4903 | |
4904 | trans = btrfs_join_transaction(root); | |
4905 | if (IS_ERR(trans)) { | |
4906 | ret = PTR_ERR(trans); | |
4907 | goto out; | |
4908 | } | |
4909 | ||
4910 | /* FIXME: check if the IDs really exist */ | |
4911 | if (sa->assign) { | |
0b246afa | 4912 | ret = btrfs_add_qgroup_relation(trans, fs_info, |
5d13a37b AJ |
4913 | sa->src, sa->dst); |
4914 | } else { | |
0b246afa | 4915 | ret = btrfs_del_qgroup_relation(trans, fs_info, |
5d13a37b AJ |
4916 | sa->src, sa->dst); |
4917 | } | |
4918 | ||
e082f563 | 4919 | /* update qgroup status and info */ |
0b246afa | 4920 | err = btrfs_run_qgroups(trans, fs_info); |
e082f563 | 4921 | if (err < 0) |
0b246afa JM |
4922 | btrfs_handle_fs_error(fs_info, err, |
4923 | "failed to update qgroup status and info"); | |
3a45bb20 | 4924 | err = btrfs_end_transaction(trans); |
5d13a37b AJ |
4925 | if (err && !ret) |
4926 | ret = err; | |
4927 | ||
4928 | out: | |
4929 | kfree(sa); | |
905b0dda MX |
4930 | drop_write: |
4931 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4932 | return ret; |
4933 | } | |
4934 | ||
905b0dda | 4935 | static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) |
5d13a37b | 4936 | { |
0b246afa JM |
4937 | struct inode *inode = file_inode(file); |
4938 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
4939 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
5d13a37b AJ |
4940 | struct btrfs_ioctl_qgroup_create_args *sa; |
4941 | struct btrfs_trans_handle *trans; | |
4942 | int ret; | |
4943 | int err; | |
4944 | ||
4945 | if (!capable(CAP_SYS_ADMIN)) | |
4946 | return -EPERM; | |
4947 | ||
905b0dda MX |
4948 | ret = mnt_want_write_file(file); |
4949 | if (ret) | |
4950 | return ret; | |
5d13a37b AJ |
4951 | |
4952 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4953 | if (IS_ERR(sa)) { |
4954 | ret = PTR_ERR(sa); | |
4955 | goto drop_write; | |
4956 | } | |
5d13a37b | 4957 | |
d86e56cf MX |
4958 | if (!sa->qgroupid) { |
4959 | ret = -EINVAL; | |
4960 | goto out; | |
4961 | } | |
4962 | ||
5d13a37b AJ |
4963 | trans = btrfs_join_transaction(root); |
4964 | if (IS_ERR(trans)) { | |
4965 | ret = PTR_ERR(trans); | |
4966 | goto out; | |
4967 | } | |
4968 | ||
4969 | /* FIXME: check if the IDs really exist */ | |
4970 | if (sa->create) { | |
0b246afa | 4971 | ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid); |
5d13a37b | 4972 | } else { |
0b246afa | 4973 | ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid); |
5d13a37b AJ |
4974 | } |
4975 | ||
3a45bb20 | 4976 | err = btrfs_end_transaction(trans); |
5d13a37b AJ |
4977 | if (err && !ret) |
4978 | ret = err; | |
4979 | ||
4980 | out: | |
4981 | kfree(sa); | |
905b0dda MX |
4982 | drop_write: |
4983 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4984 | return ret; |
4985 | } | |
4986 | ||
905b0dda | 4987 | static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) |
5d13a37b | 4988 | { |
0b246afa JM |
4989 | struct inode *inode = file_inode(file); |
4990 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
4991 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
5d13a37b AJ |
4992 | struct btrfs_ioctl_qgroup_limit_args *sa; |
4993 | struct btrfs_trans_handle *trans; | |
4994 | int ret; | |
4995 | int err; | |
4996 | u64 qgroupid; | |
4997 | ||
4998 | if (!capable(CAP_SYS_ADMIN)) | |
4999 | return -EPERM; | |
5000 | ||
905b0dda MX |
5001 | ret = mnt_want_write_file(file); |
5002 | if (ret) | |
5003 | return ret; | |
5d13a37b AJ |
5004 | |
5005 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
5006 | if (IS_ERR(sa)) { |
5007 | ret = PTR_ERR(sa); | |
5008 | goto drop_write; | |
5009 | } | |
5d13a37b AJ |
5010 | |
5011 | trans = btrfs_join_transaction(root); | |
5012 | if (IS_ERR(trans)) { | |
5013 | ret = PTR_ERR(trans); | |
5014 | goto out; | |
5015 | } | |
5016 | ||
5017 | qgroupid = sa->qgroupid; | |
5018 | if (!qgroupid) { | |
5019 | /* take the current subvol as qgroup */ | |
5020 | qgroupid = root->root_key.objectid; | |
5021 | } | |
5022 | ||
5023 | /* FIXME: check if the IDs really exist */ | |
0b246afa | 5024 | ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim); |
5d13a37b | 5025 | |
3a45bb20 | 5026 | err = btrfs_end_transaction(trans); |
5d13a37b AJ |
5027 | if (err && !ret) |
5028 | ret = err; | |
5029 | ||
5030 | out: | |
5031 | kfree(sa); | |
905b0dda MX |
5032 | drop_write: |
5033 | mnt_drop_write_file(file); | |
5d13a37b AJ |
5034 | return ret; |
5035 | } | |
5036 | ||
2f232036 JS |
5037 | static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) |
5038 | { | |
0b246afa JM |
5039 | struct inode *inode = file_inode(file); |
5040 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
2f232036 JS |
5041 | struct btrfs_ioctl_quota_rescan_args *qsa; |
5042 | int ret; | |
5043 | ||
5044 | if (!capable(CAP_SYS_ADMIN)) | |
5045 | return -EPERM; | |
5046 | ||
5047 | ret = mnt_want_write_file(file); | |
5048 | if (ret) | |
5049 | return ret; | |
5050 | ||
5051 | qsa = memdup_user(arg, sizeof(*qsa)); | |
5052 | if (IS_ERR(qsa)) { | |
5053 | ret = PTR_ERR(qsa); | |
5054 | goto drop_write; | |
5055 | } | |
5056 | ||
5057 | if (qsa->flags) { | |
5058 | ret = -EINVAL; | |
5059 | goto out; | |
5060 | } | |
5061 | ||
0b246afa | 5062 | ret = btrfs_qgroup_rescan(fs_info); |
2f232036 JS |
5063 | |
5064 | out: | |
5065 | kfree(qsa); | |
5066 | drop_write: | |
5067 | mnt_drop_write_file(file); | |
5068 | return ret; | |
5069 | } | |
5070 | ||
5071 | static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) | |
5072 | { | |
0b246afa JM |
5073 | struct inode *inode = file_inode(file); |
5074 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
2f232036 JS |
5075 | struct btrfs_ioctl_quota_rescan_args *qsa; |
5076 | int ret = 0; | |
5077 | ||
5078 | if (!capable(CAP_SYS_ADMIN)) | |
5079 | return -EPERM; | |
5080 | ||
8d2db785 | 5081 | qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); |
2f232036 JS |
5082 | if (!qsa) |
5083 | return -ENOMEM; | |
5084 | ||
0b246afa | 5085 | if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { |
2f232036 | 5086 | qsa->flags = 1; |
0b246afa | 5087 | qsa->progress = fs_info->qgroup_rescan_progress.objectid; |
2f232036 JS |
5088 | } |
5089 | ||
5090 | if (copy_to_user(arg, qsa, sizeof(*qsa))) | |
5091 | ret = -EFAULT; | |
5092 | ||
5093 | kfree(qsa); | |
5094 | return ret; | |
5095 | } | |
5096 | ||
57254b6e JS |
5097 | static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) |
5098 | { | |
0b246afa JM |
5099 | struct inode *inode = file_inode(file); |
5100 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
57254b6e JS |
5101 | |
5102 | if (!capable(CAP_SYS_ADMIN)) | |
5103 | return -EPERM; | |
5104 | ||
0b246afa | 5105 | return btrfs_qgroup_wait_for_completion(fs_info, true); |
57254b6e JS |
5106 | } |
5107 | ||
abccd00f HM |
5108 | static long _btrfs_ioctl_set_received_subvol(struct file *file, |
5109 | struct btrfs_ioctl_received_subvol_args *sa) | |
8ea05e3a | 5110 | { |
496ad9aa | 5111 | struct inode *inode = file_inode(file); |
0b246afa | 5112 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
8ea05e3a AB |
5113 | struct btrfs_root *root = BTRFS_I(inode)->root; |
5114 | struct btrfs_root_item *root_item = &root->root_item; | |
5115 | struct btrfs_trans_handle *trans; | |
c2050a45 | 5116 | struct timespec ct = current_time(inode); |
8ea05e3a | 5117 | int ret = 0; |
dd5f9615 | 5118 | int received_uuid_changed; |
8ea05e3a | 5119 | |
bd60ea0f DS |
5120 | if (!inode_owner_or_capable(inode)) |
5121 | return -EPERM; | |
5122 | ||
8ea05e3a AB |
5123 | ret = mnt_want_write_file(file); |
5124 | if (ret < 0) | |
5125 | return ret; | |
5126 | ||
0b246afa | 5127 | down_write(&fs_info->subvol_sem); |
8ea05e3a AB |
5128 | |
5129 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { | |
5130 | ret = -EINVAL; | |
5131 | goto out; | |
5132 | } | |
5133 | ||
5134 | if (btrfs_root_readonly(root)) { | |
5135 | ret = -EROFS; | |
5136 | goto out; | |
5137 | } | |
5138 | ||
dd5f9615 SB |
5139 | /* |
5140 | * 1 - root item | |
5141 | * 2 - uuid items (received uuid + subvol uuid) | |
5142 | */ | |
5143 | trans = btrfs_start_transaction(root, 3); | |
8ea05e3a AB |
5144 | if (IS_ERR(trans)) { |
5145 | ret = PTR_ERR(trans); | |
5146 | trans = NULL; | |
5147 | goto out; | |
5148 | } | |
5149 | ||
5150 | sa->rtransid = trans->transid; | |
5151 | sa->rtime.sec = ct.tv_sec; | |
5152 | sa->rtime.nsec = ct.tv_nsec; | |
5153 | ||
dd5f9615 SB |
5154 | received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, |
5155 | BTRFS_UUID_SIZE); | |
5156 | if (received_uuid_changed && | |
5157 | !btrfs_is_empty_uuid(root_item->received_uuid)) | |
0b246afa | 5158 | btrfs_uuid_tree_rem(trans, fs_info, root_item->received_uuid, |
dd5f9615 SB |
5159 | BTRFS_UUID_KEY_RECEIVED_SUBVOL, |
5160 | root->root_key.objectid); | |
8ea05e3a AB |
5161 | memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); |
5162 | btrfs_set_root_stransid(root_item, sa->stransid); | |
5163 | btrfs_set_root_rtransid(root_item, sa->rtransid); | |
3cae210f QW |
5164 | btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); |
5165 | btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); | |
5166 | btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); | |
5167 | btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); | |
8ea05e3a | 5168 | |
0b246afa | 5169 | ret = btrfs_update_root(trans, fs_info->tree_root, |
8ea05e3a AB |
5170 | &root->root_key, &root->root_item); |
5171 | if (ret < 0) { | |
3a45bb20 | 5172 | btrfs_end_transaction(trans); |
8ea05e3a | 5173 | goto out; |
dd5f9615 SB |
5174 | } |
5175 | if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { | |
0b246afa | 5176 | ret = btrfs_uuid_tree_add(trans, fs_info, sa->uuid, |
dd5f9615 SB |
5177 | BTRFS_UUID_KEY_RECEIVED_SUBVOL, |
5178 | root->root_key.objectid); | |
5179 | if (ret < 0 && ret != -EEXIST) { | |
66642832 | 5180 | btrfs_abort_transaction(trans, ret); |
8ea05e3a | 5181 | goto out; |
dd5f9615 SB |
5182 | } |
5183 | } | |
3a45bb20 | 5184 | ret = btrfs_commit_transaction(trans); |
dd5f9615 | 5185 | if (ret < 0) { |
66642832 | 5186 | btrfs_abort_transaction(trans, ret); |
dd5f9615 | 5187 | goto out; |
8ea05e3a AB |
5188 | } |
5189 | ||
abccd00f | 5190 | out: |
0b246afa | 5191 | up_write(&fs_info->subvol_sem); |
abccd00f HM |
5192 | mnt_drop_write_file(file); |
5193 | return ret; | |
5194 | } | |
5195 | ||
5196 | #ifdef CONFIG_64BIT | |
5197 | static long btrfs_ioctl_set_received_subvol_32(struct file *file, | |
5198 | void __user *arg) | |
5199 | { | |
5200 | struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; | |
5201 | struct btrfs_ioctl_received_subvol_args *args64 = NULL; | |
5202 | int ret = 0; | |
5203 | ||
5204 | args32 = memdup_user(arg, sizeof(*args32)); | |
7b9ea627 SV |
5205 | if (IS_ERR(args32)) |
5206 | return PTR_ERR(args32); | |
abccd00f | 5207 | |
8d2db785 | 5208 | args64 = kmalloc(sizeof(*args64), GFP_KERNEL); |
84dbeb87 DC |
5209 | if (!args64) { |
5210 | ret = -ENOMEM; | |
abccd00f HM |
5211 | goto out; |
5212 | } | |
5213 | ||
5214 | memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); | |
5215 | args64->stransid = args32->stransid; | |
5216 | args64->rtransid = args32->rtransid; | |
5217 | args64->stime.sec = args32->stime.sec; | |
5218 | args64->stime.nsec = args32->stime.nsec; | |
5219 | args64->rtime.sec = args32->rtime.sec; | |
5220 | args64->rtime.nsec = args32->rtime.nsec; | |
5221 | args64->flags = args32->flags; | |
5222 | ||
5223 | ret = _btrfs_ioctl_set_received_subvol(file, args64); | |
5224 | if (ret) | |
5225 | goto out; | |
5226 | ||
5227 | memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); | |
5228 | args32->stransid = args64->stransid; | |
5229 | args32->rtransid = args64->rtransid; | |
5230 | args32->stime.sec = args64->stime.sec; | |
5231 | args32->stime.nsec = args64->stime.nsec; | |
5232 | args32->rtime.sec = args64->rtime.sec; | |
5233 | args32->rtime.nsec = args64->rtime.nsec; | |
5234 | args32->flags = args64->flags; | |
5235 | ||
5236 | ret = copy_to_user(arg, args32, sizeof(*args32)); | |
5237 | if (ret) | |
5238 | ret = -EFAULT; | |
5239 | ||
5240 | out: | |
5241 | kfree(args32); | |
5242 | kfree(args64); | |
5243 | return ret; | |
5244 | } | |
5245 | #endif | |
5246 | ||
5247 | static long btrfs_ioctl_set_received_subvol(struct file *file, | |
5248 | void __user *arg) | |
5249 | { | |
5250 | struct btrfs_ioctl_received_subvol_args *sa = NULL; | |
5251 | int ret = 0; | |
5252 | ||
5253 | sa = memdup_user(arg, sizeof(*sa)); | |
7b9ea627 SV |
5254 | if (IS_ERR(sa)) |
5255 | return PTR_ERR(sa); | |
abccd00f HM |
5256 | |
5257 | ret = _btrfs_ioctl_set_received_subvol(file, sa); | |
5258 | ||
5259 | if (ret) | |
5260 | goto out; | |
5261 | ||
8ea05e3a AB |
5262 | ret = copy_to_user(arg, sa, sizeof(*sa)); |
5263 | if (ret) | |
5264 | ret = -EFAULT; | |
5265 | ||
5266 | out: | |
5267 | kfree(sa); | |
8ea05e3a AB |
5268 | return ret; |
5269 | } | |
5270 | ||
867ab667 | 5271 | static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) |
5272 | { | |
0b246afa JM |
5273 | struct inode *inode = file_inode(file); |
5274 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
a1b83ac5 | 5275 | size_t len; |
867ab667 | 5276 | int ret; |
a1b83ac5 AJ |
5277 | char label[BTRFS_LABEL_SIZE]; |
5278 | ||
0b246afa JM |
5279 | spin_lock(&fs_info->super_lock); |
5280 | memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); | |
5281 | spin_unlock(&fs_info->super_lock); | |
a1b83ac5 AJ |
5282 | |
5283 | len = strnlen(label, BTRFS_LABEL_SIZE); | |
867ab667 | 5284 | |
5285 | if (len == BTRFS_LABEL_SIZE) { | |
0b246afa JM |
5286 | btrfs_warn(fs_info, |
5287 | "label is too long, return the first %zu bytes", | |
5288 | --len); | |
867ab667 | 5289 | } |
5290 | ||
867ab667 | 5291 | ret = copy_to_user(arg, label, len); |
867ab667 | 5292 | |
5293 | return ret ? -EFAULT : 0; | |
5294 | } | |
5295 | ||
a8bfd4ab | 5296 | static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) |
5297 | { | |
0b246afa JM |
5298 | struct inode *inode = file_inode(file); |
5299 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
5300 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
5301 | struct btrfs_super_block *super_block = fs_info->super_copy; | |
a8bfd4ab | 5302 | struct btrfs_trans_handle *trans; |
5303 | char label[BTRFS_LABEL_SIZE]; | |
5304 | int ret; | |
5305 | ||
5306 | if (!capable(CAP_SYS_ADMIN)) | |
5307 | return -EPERM; | |
5308 | ||
5309 | if (copy_from_user(label, arg, sizeof(label))) | |
5310 | return -EFAULT; | |
5311 | ||
5312 | if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { | |
0b246afa | 5313 | btrfs_err(fs_info, |
5d163e0e JM |
5314 | "unable to set label with more than %d bytes", |
5315 | BTRFS_LABEL_SIZE - 1); | |
a8bfd4ab | 5316 | return -EINVAL; |
5317 | } | |
5318 | ||
5319 | ret = mnt_want_write_file(file); | |
5320 | if (ret) | |
5321 | return ret; | |
5322 | ||
a8bfd4ab | 5323 | trans = btrfs_start_transaction(root, 0); |
5324 | if (IS_ERR(trans)) { | |
5325 | ret = PTR_ERR(trans); | |
5326 | goto out_unlock; | |
5327 | } | |
5328 | ||
0b246afa | 5329 | spin_lock(&fs_info->super_lock); |
a8bfd4ab | 5330 | strcpy(super_block->label, label); |
0b246afa | 5331 | spin_unlock(&fs_info->super_lock); |
3a45bb20 | 5332 | ret = btrfs_commit_transaction(trans); |
a8bfd4ab | 5333 | |
5334 | out_unlock: | |
a8bfd4ab | 5335 | mnt_drop_write_file(file); |
5336 | return ret; | |
5337 | } | |
5338 | ||
2eaa055f JM |
5339 | #define INIT_FEATURE_FLAGS(suffix) \ |
5340 | { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ | |
5341 | .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ | |
5342 | .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } | |
5343 | ||
d5131b65 | 5344 | int btrfs_ioctl_get_supported_features(void __user *arg) |
2eaa055f | 5345 | { |
4d4ab6d6 | 5346 | static const struct btrfs_ioctl_feature_flags features[3] = { |
2eaa055f JM |
5347 | INIT_FEATURE_FLAGS(SUPP), |
5348 | INIT_FEATURE_FLAGS(SAFE_SET), | |
5349 | INIT_FEATURE_FLAGS(SAFE_CLEAR) | |
5350 | }; | |
5351 | ||
5352 | if (copy_to_user(arg, &features, sizeof(features))) | |
5353 | return -EFAULT; | |
5354 | ||
5355 | return 0; | |
5356 | } | |
5357 | ||
5358 | static int btrfs_ioctl_get_features(struct file *file, void __user *arg) | |
5359 | { | |
0b246afa JM |
5360 | struct inode *inode = file_inode(file); |
5361 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
5362 | struct btrfs_super_block *super_block = fs_info->super_copy; | |
2eaa055f JM |
5363 | struct btrfs_ioctl_feature_flags features; |
5364 | ||
5365 | features.compat_flags = btrfs_super_compat_flags(super_block); | |
5366 | features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); | |
5367 | features.incompat_flags = btrfs_super_incompat_flags(super_block); | |
5368 | ||
5369 | if (copy_to_user(arg, &features, sizeof(features))) | |
5370 | return -EFAULT; | |
5371 | ||
5372 | return 0; | |
5373 | } | |
5374 | ||
2ff7e61e | 5375 | static int check_feature_bits(struct btrfs_fs_info *fs_info, |
3b02a68a | 5376 | enum btrfs_feature_set set, |
2eaa055f JM |
5377 | u64 change_mask, u64 flags, u64 supported_flags, |
5378 | u64 safe_set, u64 safe_clear) | |
5379 | { | |
3b02a68a JM |
5380 | const char *type = btrfs_feature_set_names[set]; |
5381 | char *names; | |
2eaa055f JM |
5382 | u64 disallowed, unsupported; |
5383 | u64 set_mask = flags & change_mask; | |
5384 | u64 clear_mask = ~flags & change_mask; | |
5385 | ||
5386 | unsupported = set_mask & ~supported_flags; | |
5387 | if (unsupported) { | |
3b02a68a JM |
5388 | names = btrfs_printable_features(set, unsupported); |
5389 | if (names) { | |
0b246afa JM |
5390 | btrfs_warn(fs_info, |
5391 | "this kernel does not support the %s feature bit%s", | |
5392 | names, strchr(names, ',') ? "s" : ""); | |
3b02a68a JM |
5393 | kfree(names); |
5394 | } else | |
0b246afa JM |
5395 | btrfs_warn(fs_info, |
5396 | "this kernel does not support %s bits 0x%llx", | |
5397 | type, unsupported); | |
2eaa055f JM |
5398 | return -EOPNOTSUPP; |
5399 | } | |
5400 | ||
5401 | disallowed = set_mask & ~safe_set; | |
5402 | if (disallowed) { | |
3b02a68a JM |
5403 | names = btrfs_printable_features(set, disallowed); |
5404 | if (names) { | |
0b246afa JM |
5405 | btrfs_warn(fs_info, |
5406 | "can't set the %s feature bit%s while mounted", | |
5407 | names, strchr(names, ',') ? "s" : ""); | |
3b02a68a JM |
5408 | kfree(names); |
5409 | } else | |
0b246afa JM |
5410 | btrfs_warn(fs_info, |
5411 | "can't set %s bits 0x%llx while mounted", | |
5412 | type, disallowed); | |
2eaa055f JM |
5413 | return -EPERM; |
5414 | } | |
5415 | ||
5416 | disallowed = clear_mask & ~safe_clear; | |
5417 | if (disallowed) { | |
3b02a68a JM |
5418 | names = btrfs_printable_features(set, disallowed); |
5419 | if (names) { | |
0b246afa JM |
5420 | btrfs_warn(fs_info, |
5421 | "can't clear the %s feature bit%s while mounted", | |
5422 | names, strchr(names, ',') ? "s" : ""); | |
3b02a68a JM |
5423 | kfree(names); |
5424 | } else | |
0b246afa JM |
5425 | btrfs_warn(fs_info, |
5426 | "can't clear %s bits 0x%llx while mounted", | |
5427 | type, disallowed); | |
2eaa055f JM |
5428 | return -EPERM; |
5429 | } | |
5430 | ||
5431 | return 0; | |
5432 | } | |
5433 | ||
2ff7e61e JM |
5434 | #define check_feature(fs_info, change_mask, flags, mask_base) \ |
5435 | check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ | |
2eaa055f JM |
5436 | BTRFS_FEATURE_ ## mask_base ## _SUPP, \ |
5437 | BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ | |
5438 | BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) | |
5439 | ||
5440 | static int btrfs_ioctl_set_features(struct file *file, void __user *arg) | |
5441 | { | |
0b246afa JM |
5442 | struct inode *inode = file_inode(file); |
5443 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
5444 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
5445 | struct btrfs_super_block *super_block = fs_info->super_copy; | |
2eaa055f JM |
5446 | struct btrfs_ioctl_feature_flags flags[2]; |
5447 | struct btrfs_trans_handle *trans; | |
5448 | u64 newflags; | |
5449 | int ret; | |
5450 | ||
5451 | if (!capable(CAP_SYS_ADMIN)) | |
5452 | return -EPERM; | |
5453 | ||
5454 | if (copy_from_user(flags, arg, sizeof(flags))) | |
5455 | return -EFAULT; | |
5456 | ||
5457 | /* Nothing to do */ | |
5458 | if (!flags[0].compat_flags && !flags[0].compat_ro_flags && | |
5459 | !flags[0].incompat_flags) | |
5460 | return 0; | |
5461 | ||
2ff7e61e | 5462 | ret = check_feature(fs_info, flags[0].compat_flags, |
2eaa055f JM |
5463 | flags[1].compat_flags, COMPAT); |
5464 | if (ret) | |
5465 | return ret; | |
5466 | ||
2ff7e61e | 5467 | ret = check_feature(fs_info, flags[0].compat_ro_flags, |
2eaa055f JM |
5468 | flags[1].compat_ro_flags, COMPAT_RO); |
5469 | if (ret) | |
5470 | return ret; | |
5471 | ||
2ff7e61e | 5472 | ret = check_feature(fs_info, flags[0].incompat_flags, |
2eaa055f JM |
5473 | flags[1].incompat_flags, INCOMPAT); |
5474 | if (ret) | |
5475 | return ret; | |
5476 | ||
7ab19625 DS |
5477 | ret = mnt_want_write_file(file); |
5478 | if (ret) | |
5479 | return ret; | |
5480 | ||
8051aa1a | 5481 | trans = btrfs_start_transaction(root, 0); |
7ab19625 DS |
5482 | if (IS_ERR(trans)) { |
5483 | ret = PTR_ERR(trans); | |
5484 | goto out_drop_write; | |
5485 | } | |
2eaa055f | 5486 | |
0b246afa | 5487 | spin_lock(&fs_info->super_lock); |
2eaa055f JM |
5488 | newflags = btrfs_super_compat_flags(super_block); |
5489 | newflags |= flags[0].compat_flags & flags[1].compat_flags; | |
5490 | newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); | |
5491 | btrfs_set_super_compat_flags(super_block, newflags); | |
5492 | ||
5493 | newflags = btrfs_super_compat_ro_flags(super_block); | |
5494 | newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; | |
5495 | newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); | |
5496 | btrfs_set_super_compat_ro_flags(super_block, newflags); | |
5497 | ||
5498 | newflags = btrfs_super_incompat_flags(super_block); | |
5499 | newflags |= flags[0].incompat_flags & flags[1].incompat_flags; | |
5500 | newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); | |
5501 | btrfs_set_super_incompat_flags(super_block, newflags); | |
0b246afa | 5502 | spin_unlock(&fs_info->super_lock); |
2eaa055f | 5503 | |
3a45bb20 | 5504 | ret = btrfs_commit_transaction(trans); |
7ab19625 DS |
5505 | out_drop_write: |
5506 | mnt_drop_write_file(file); | |
5507 | ||
5508 | return ret; | |
2eaa055f JM |
5509 | } |
5510 | ||
f46b5a66 CH |
5511 | long btrfs_ioctl(struct file *file, unsigned int |
5512 | cmd, unsigned long arg) | |
5513 | { | |
0b246afa JM |
5514 | struct inode *inode = file_inode(file); |
5515 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
5516 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
4bcabaa3 | 5517 | void __user *argp = (void __user *)arg; |
f46b5a66 CH |
5518 | |
5519 | switch (cmd) { | |
6cbff00f CH |
5520 | case FS_IOC_GETFLAGS: |
5521 | return btrfs_ioctl_getflags(file, argp); | |
5522 | case FS_IOC_SETFLAGS: | |
5523 | return btrfs_ioctl_setflags(file, argp); | |
5524 | case FS_IOC_GETVERSION: | |
5525 | return btrfs_ioctl_getversion(file, argp); | |
f7039b1d LD |
5526 | case FITRIM: |
5527 | return btrfs_ioctl_fitrim(file, argp); | |
f46b5a66 | 5528 | case BTRFS_IOC_SNAP_CREATE: |
fa0d2b9b | 5529 | return btrfs_ioctl_snap_create(file, argp, 0); |
fdfb1e4f | 5530 | case BTRFS_IOC_SNAP_CREATE_V2: |
fa0d2b9b | 5531 | return btrfs_ioctl_snap_create_v2(file, argp, 0); |
3de4586c | 5532 | case BTRFS_IOC_SUBVOL_CREATE: |
fa0d2b9b | 5533 | return btrfs_ioctl_snap_create(file, argp, 1); |
6f72c7e2 AJ |
5534 | case BTRFS_IOC_SUBVOL_CREATE_V2: |
5535 | return btrfs_ioctl_snap_create_v2(file, argp, 1); | |
76dda93c YZ |
5536 | case BTRFS_IOC_SNAP_DESTROY: |
5537 | return btrfs_ioctl_snap_destroy(file, argp); | |
0caa102d LZ |
5538 | case BTRFS_IOC_SUBVOL_GETFLAGS: |
5539 | return btrfs_ioctl_subvol_getflags(file, argp); | |
5540 | case BTRFS_IOC_SUBVOL_SETFLAGS: | |
5541 | return btrfs_ioctl_subvol_setflags(file, argp); | |
6ef5ed0d JB |
5542 | case BTRFS_IOC_DEFAULT_SUBVOL: |
5543 | return btrfs_ioctl_default_subvol(file, argp); | |
f46b5a66 | 5544 | case BTRFS_IOC_DEFRAG: |
1e701a32 CM |
5545 | return btrfs_ioctl_defrag(file, NULL); |
5546 | case BTRFS_IOC_DEFRAG_RANGE: | |
5547 | return btrfs_ioctl_defrag(file, argp); | |
f46b5a66 | 5548 | case BTRFS_IOC_RESIZE: |
198605a8 | 5549 | return btrfs_ioctl_resize(file, argp); |
f46b5a66 | 5550 | case BTRFS_IOC_ADD_DEV: |
2ff7e61e | 5551 | return btrfs_ioctl_add_dev(fs_info, argp); |
f46b5a66 | 5552 | case BTRFS_IOC_RM_DEV: |
da24927b | 5553 | return btrfs_ioctl_rm_dev(file, argp); |
6b526ed7 AJ |
5554 | case BTRFS_IOC_RM_DEV_V2: |
5555 | return btrfs_ioctl_rm_dev_v2(file, argp); | |
475f6387 | 5556 | case BTRFS_IOC_FS_INFO: |
2ff7e61e | 5557 | return btrfs_ioctl_fs_info(fs_info, argp); |
475f6387 | 5558 | case BTRFS_IOC_DEV_INFO: |
2ff7e61e | 5559 | return btrfs_ioctl_dev_info(fs_info, argp); |
f46b5a66 | 5560 | case BTRFS_IOC_BALANCE: |
9ba1f6e4 | 5561 | return btrfs_ioctl_balance(file, NULL); |
f46b5a66 CH |
5562 | case BTRFS_IOC_TRANS_START: |
5563 | return btrfs_ioctl_trans_start(file); | |
5564 | case BTRFS_IOC_TRANS_END: | |
5565 | return btrfs_ioctl_trans_end(file); | |
ac8e9819 CM |
5566 | case BTRFS_IOC_TREE_SEARCH: |
5567 | return btrfs_ioctl_tree_search(file, argp); | |
cc68a8a5 GH |
5568 | case BTRFS_IOC_TREE_SEARCH_V2: |
5569 | return btrfs_ioctl_tree_search_v2(file, argp); | |
ac8e9819 CM |
5570 | case BTRFS_IOC_INO_LOOKUP: |
5571 | return btrfs_ioctl_ino_lookup(file, argp); | |
d7728c96 JS |
5572 | case BTRFS_IOC_INO_PATHS: |
5573 | return btrfs_ioctl_ino_to_path(root, argp); | |
5574 | case BTRFS_IOC_LOGICAL_INO: | |
2ff7e61e | 5575 | return btrfs_ioctl_logical_to_ino(fs_info, argp); |
1406e432 | 5576 | case BTRFS_IOC_SPACE_INFO: |
2ff7e61e | 5577 | return btrfs_ioctl_space_info(fs_info, argp); |
9b199859 FDBM |
5578 | case BTRFS_IOC_SYNC: { |
5579 | int ret; | |
5580 | ||
0b246afa | 5581 | ret = btrfs_start_delalloc_roots(fs_info, 0, -1); |
9b199859 FDBM |
5582 | if (ret) |
5583 | return ret; | |
0b246afa | 5584 | ret = btrfs_sync_fs(inode->i_sb, 1); |
2fad4e83 DS |
5585 | /* |
5586 | * The transaction thread may want to do more work, | |
01327610 | 5587 | * namely it pokes the cleaner kthread that will start |
2fad4e83 DS |
5588 | * processing uncleaned subvols. |
5589 | */ | |
0b246afa | 5590 | wake_up_process(fs_info->transaction_kthread); |
9b199859 FDBM |
5591 | return ret; |
5592 | } | |
46204592 | 5593 | case BTRFS_IOC_START_SYNC: |
9a8c28be | 5594 | return btrfs_ioctl_start_sync(root, argp); |
46204592 | 5595 | case BTRFS_IOC_WAIT_SYNC: |
2ff7e61e | 5596 | return btrfs_ioctl_wait_sync(fs_info, argp); |
475f6387 | 5597 | case BTRFS_IOC_SCRUB: |
b8e95489 | 5598 | return btrfs_ioctl_scrub(file, argp); |
475f6387 | 5599 | case BTRFS_IOC_SCRUB_CANCEL: |
2ff7e61e | 5600 | return btrfs_ioctl_scrub_cancel(fs_info); |
475f6387 | 5601 | case BTRFS_IOC_SCRUB_PROGRESS: |
2ff7e61e | 5602 | return btrfs_ioctl_scrub_progress(fs_info, argp); |
c9e9f97b | 5603 | case BTRFS_IOC_BALANCE_V2: |
9ba1f6e4 | 5604 | return btrfs_ioctl_balance(file, argp); |
837d5b6e | 5605 | case BTRFS_IOC_BALANCE_CTL: |
2ff7e61e | 5606 | return btrfs_ioctl_balance_ctl(fs_info, arg); |
19a39dce | 5607 | case BTRFS_IOC_BALANCE_PROGRESS: |
2ff7e61e | 5608 | return btrfs_ioctl_balance_progress(fs_info, argp); |
8ea05e3a AB |
5609 | case BTRFS_IOC_SET_RECEIVED_SUBVOL: |
5610 | return btrfs_ioctl_set_received_subvol(file, argp); | |
abccd00f HM |
5611 | #ifdef CONFIG_64BIT |
5612 | case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: | |
5613 | return btrfs_ioctl_set_received_subvol_32(file, argp); | |
5614 | #endif | |
31db9f7c AB |
5615 | case BTRFS_IOC_SEND: |
5616 | return btrfs_ioctl_send(file, argp); | |
c11d2c23 | 5617 | case BTRFS_IOC_GET_DEV_STATS: |
2ff7e61e | 5618 | return btrfs_ioctl_get_dev_stats(fs_info, argp); |
5d13a37b | 5619 | case BTRFS_IOC_QUOTA_CTL: |
905b0dda | 5620 | return btrfs_ioctl_quota_ctl(file, argp); |
5d13a37b | 5621 | case BTRFS_IOC_QGROUP_ASSIGN: |
905b0dda | 5622 | return btrfs_ioctl_qgroup_assign(file, argp); |
5d13a37b | 5623 | case BTRFS_IOC_QGROUP_CREATE: |
905b0dda | 5624 | return btrfs_ioctl_qgroup_create(file, argp); |
5d13a37b | 5625 | case BTRFS_IOC_QGROUP_LIMIT: |
905b0dda | 5626 | return btrfs_ioctl_qgroup_limit(file, argp); |
2f232036 JS |
5627 | case BTRFS_IOC_QUOTA_RESCAN: |
5628 | return btrfs_ioctl_quota_rescan(file, argp); | |
5629 | case BTRFS_IOC_QUOTA_RESCAN_STATUS: | |
5630 | return btrfs_ioctl_quota_rescan_status(file, argp); | |
57254b6e JS |
5631 | case BTRFS_IOC_QUOTA_RESCAN_WAIT: |
5632 | return btrfs_ioctl_quota_rescan_wait(file, argp); | |
3f6bcfbd | 5633 | case BTRFS_IOC_DEV_REPLACE: |
2ff7e61e | 5634 | return btrfs_ioctl_dev_replace(fs_info, argp); |
867ab667 | 5635 | case BTRFS_IOC_GET_FSLABEL: |
5636 | return btrfs_ioctl_get_fslabel(file, argp); | |
a8bfd4ab | 5637 | case BTRFS_IOC_SET_FSLABEL: |
5638 | return btrfs_ioctl_set_fslabel(file, argp); | |
2eaa055f | 5639 | case BTRFS_IOC_GET_SUPPORTED_FEATURES: |
d5131b65 | 5640 | return btrfs_ioctl_get_supported_features(argp); |
2eaa055f JM |
5641 | case BTRFS_IOC_GET_FEATURES: |
5642 | return btrfs_ioctl_get_features(file, argp); | |
5643 | case BTRFS_IOC_SET_FEATURES: | |
5644 | return btrfs_ioctl_set_features(file, argp); | |
f46b5a66 CH |
5645 | } |
5646 | ||
5647 | return -ENOTTY; | |
5648 | } | |
4c63c245 LD |
5649 | |
5650 | #ifdef CONFIG_COMPAT | |
5651 | long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) | |
5652 | { | |
2a362249 JM |
5653 | /* |
5654 | * These all access 32-bit values anyway so no further | |
5655 | * handling is necessary. | |
5656 | */ | |
4c63c245 LD |
5657 | switch (cmd) { |
5658 | case FS_IOC32_GETFLAGS: | |
5659 | cmd = FS_IOC_GETFLAGS; | |
5660 | break; | |
5661 | case FS_IOC32_SETFLAGS: | |
5662 | cmd = FS_IOC_SETFLAGS; | |
5663 | break; | |
5664 | case FS_IOC32_GETVERSION: | |
5665 | cmd = FS_IOC_GETVERSION; | |
5666 | break; | |
4c63c245 LD |
5667 | } |
5668 | ||
5669 | return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); | |
5670 | } | |
5671 | #endif |