Commit | Line | Data |
---|---|---|
f46b5a66 CH |
1 | /* |
2 | * Copyright (C) 2007 Oracle. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | ||
19 | #include <linux/kernel.h> | |
20 | #include <linux/bio.h> | |
21 | #include <linux/buffer_head.h> | |
22 | #include <linux/file.h> | |
23 | #include <linux/fs.h> | |
cb8e7090 | 24 | #include <linux/fsnotify.h> |
f46b5a66 CH |
25 | #include <linux/pagemap.h> |
26 | #include <linux/highmem.h> | |
27 | #include <linux/time.h> | |
28 | #include <linux/init.h> | |
29 | #include <linux/string.h> | |
f46b5a66 | 30 | #include <linux/backing-dev.h> |
cb8e7090 | 31 | #include <linux/mount.h> |
f46b5a66 | 32 | #include <linux/mpage.h> |
cb8e7090 | 33 | #include <linux/namei.h> |
f46b5a66 CH |
34 | #include <linux/swap.h> |
35 | #include <linux/writeback.h> | |
36 | #include <linux/statfs.h> | |
37 | #include <linux/compat.h> | |
38 | #include <linux/bit_spinlock.h> | |
cb8e7090 | 39 | #include <linux/security.h> |
f46b5a66 | 40 | #include <linux/xattr.h> |
7ea394f1 | 41 | #include <linux/vmalloc.h> |
5a0e3ad6 | 42 | #include <linux/slab.h> |
f7039b1d | 43 | #include <linux/blkdev.h> |
8ea05e3a | 44 | #include <linux/uuid.h> |
55e301fd | 45 | #include <linux/btrfs.h> |
416161db | 46 | #include <linux/uaccess.h> |
f46b5a66 CH |
47 | #include "ctree.h" |
48 | #include "disk-io.h" | |
49 | #include "transaction.h" | |
50 | #include "btrfs_inode.h" | |
f46b5a66 CH |
51 | #include "print-tree.h" |
52 | #include "volumes.h" | |
925baedd | 53 | #include "locking.h" |
581bb050 | 54 | #include "inode-map.h" |
d7728c96 | 55 | #include "backref.h" |
606686ee | 56 | #include "rcu-string.h" |
31db9f7c | 57 | #include "send.h" |
3f6bcfbd | 58 | #include "dev-replace.h" |
63541927 | 59 | #include "props.h" |
3b02a68a | 60 | #include "sysfs.h" |
fcebe456 | 61 | #include "qgroup.h" |
f46b5a66 | 62 | |
abccd00f HM |
63 | #ifdef CONFIG_64BIT |
64 | /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI | |
65 | * structures are incorrect, as the timespec structure from userspace | |
66 | * is 4 bytes too small. We define these alternatives here to teach | |
67 | * the kernel about the 32-bit struct packing. | |
68 | */ | |
69 | struct btrfs_ioctl_timespec_32 { | |
70 | __u64 sec; | |
71 | __u32 nsec; | |
72 | } __attribute__ ((__packed__)); | |
73 | ||
74 | struct btrfs_ioctl_received_subvol_args_32 { | |
75 | char uuid[BTRFS_UUID_SIZE]; /* in */ | |
76 | __u64 stransid; /* in */ | |
77 | __u64 rtransid; /* out */ | |
78 | struct btrfs_ioctl_timespec_32 stime; /* in */ | |
79 | struct btrfs_ioctl_timespec_32 rtime; /* out */ | |
80 | __u64 flags; /* in */ | |
81 | __u64 reserved[16]; /* in */ | |
82 | } __attribute__ ((__packed__)); | |
83 | ||
84 | #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ | |
85 | struct btrfs_ioctl_received_subvol_args_32) | |
86 | #endif | |
87 | ||
88 | ||
416161db MF |
89 | static int btrfs_clone(struct inode *src, struct inode *inode, |
90 | u64 off, u64 olen, u64 olen_aligned, u64 destoff); | |
91 | ||
6cbff00f CH |
92 | /* Mask out flags that are inappropriate for the given type of inode. */ |
93 | static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags) | |
94 | { | |
95 | if (S_ISDIR(mode)) | |
96 | return flags; | |
97 | else if (S_ISREG(mode)) | |
98 | return flags & ~FS_DIRSYNC_FL; | |
99 | else | |
100 | return flags & (FS_NODUMP_FL | FS_NOATIME_FL); | |
101 | } | |
102 | ||
103 | /* | |
104 | * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl. | |
105 | */ | |
106 | static unsigned int btrfs_flags_to_ioctl(unsigned int flags) | |
107 | { | |
108 | unsigned int iflags = 0; | |
109 | ||
110 | if (flags & BTRFS_INODE_SYNC) | |
111 | iflags |= FS_SYNC_FL; | |
112 | if (flags & BTRFS_INODE_IMMUTABLE) | |
113 | iflags |= FS_IMMUTABLE_FL; | |
114 | if (flags & BTRFS_INODE_APPEND) | |
115 | iflags |= FS_APPEND_FL; | |
116 | if (flags & BTRFS_INODE_NODUMP) | |
117 | iflags |= FS_NODUMP_FL; | |
118 | if (flags & BTRFS_INODE_NOATIME) | |
119 | iflags |= FS_NOATIME_FL; | |
120 | if (flags & BTRFS_INODE_DIRSYNC) | |
121 | iflags |= FS_DIRSYNC_FL; | |
d0092bdd LZ |
122 | if (flags & BTRFS_INODE_NODATACOW) |
123 | iflags |= FS_NOCOW_FL; | |
124 | ||
125 | if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS)) | |
126 | iflags |= FS_COMPR_FL; | |
127 | else if (flags & BTRFS_INODE_NOCOMPRESS) | |
128 | iflags |= FS_NOCOMP_FL; | |
6cbff00f CH |
129 | |
130 | return iflags; | |
131 | } | |
132 | ||
133 | /* | |
134 | * Update inode->i_flags based on the btrfs internal flags. | |
135 | */ | |
136 | void btrfs_update_iflags(struct inode *inode) | |
137 | { | |
138 | struct btrfs_inode *ip = BTRFS_I(inode); | |
3cc79392 | 139 | unsigned int new_fl = 0; |
6cbff00f CH |
140 | |
141 | if (ip->flags & BTRFS_INODE_SYNC) | |
3cc79392 | 142 | new_fl |= S_SYNC; |
6cbff00f | 143 | if (ip->flags & BTRFS_INODE_IMMUTABLE) |
3cc79392 | 144 | new_fl |= S_IMMUTABLE; |
6cbff00f | 145 | if (ip->flags & BTRFS_INODE_APPEND) |
3cc79392 | 146 | new_fl |= S_APPEND; |
6cbff00f | 147 | if (ip->flags & BTRFS_INODE_NOATIME) |
3cc79392 | 148 | new_fl |= S_NOATIME; |
6cbff00f | 149 | if (ip->flags & BTRFS_INODE_DIRSYNC) |
3cc79392 FM |
150 | new_fl |= S_DIRSYNC; |
151 | ||
152 | set_mask_bits(&inode->i_flags, | |
153 | S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC, | |
154 | new_fl); | |
6cbff00f CH |
155 | } |
156 | ||
157 | /* | |
158 | * Inherit flags from the parent inode. | |
159 | * | |
e27425d6 | 160 | * Currently only the compression flags and the cow flags are inherited. |
6cbff00f CH |
161 | */ |
162 | void btrfs_inherit_iflags(struct inode *inode, struct inode *dir) | |
163 | { | |
0b4dcea5 CM |
164 | unsigned int flags; |
165 | ||
166 | if (!dir) | |
167 | return; | |
168 | ||
169 | flags = BTRFS_I(dir)->flags; | |
6cbff00f | 170 | |
e27425d6 JB |
171 | if (flags & BTRFS_INODE_NOCOMPRESS) { |
172 | BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS; | |
173 | BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS; | |
174 | } else if (flags & BTRFS_INODE_COMPRESS) { | |
175 | BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS; | |
176 | BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS; | |
177 | } | |
178 | ||
213490b3 | 179 | if (flags & BTRFS_INODE_NODATACOW) { |
e27425d6 | 180 | BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW; |
213490b3 LB |
181 | if (S_ISREG(inode->i_mode)) |
182 | BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM; | |
183 | } | |
6cbff00f | 184 | |
6cbff00f CH |
185 | btrfs_update_iflags(inode); |
186 | } | |
187 | ||
188 | static int btrfs_ioctl_getflags(struct file *file, void __user *arg) | |
189 | { | |
496ad9aa | 190 | struct btrfs_inode *ip = BTRFS_I(file_inode(file)); |
6cbff00f CH |
191 | unsigned int flags = btrfs_flags_to_ioctl(ip->flags); |
192 | ||
193 | if (copy_to_user(arg, &flags, sizeof(flags))) | |
194 | return -EFAULT; | |
195 | return 0; | |
196 | } | |
197 | ||
75e7cb7f LB |
198 | static int check_flags(unsigned int flags) |
199 | { | |
200 | if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ | |
201 | FS_NOATIME_FL | FS_NODUMP_FL | \ | |
202 | FS_SYNC_FL | FS_DIRSYNC_FL | \ | |
e1e8fb6a LZ |
203 | FS_NOCOMP_FL | FS_COMPR_FL | |
204 | FS_NOCOW_FL)) | |
75e7cb7f LB |
205 | return -EOPNOTSUPP; |
206 | ||
207 | if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) | |
208 | return -EINVAL; | |
209 | ||
75e7cb7f LB |
210 | return 0; |
211 | } | |
212 | ||
6cbff00f CH |
213 | static int btrfs_ioctl_setflags(struct file *file, void __user *arg) |
214 | { | |
496ad9aa | 215 | struct inode *inode = file_inode(file); |
6cbff00f CH |
216 | struct btrfs_inode *ip = BTRFS_I(inode); |
217 | struct btrfs_root *root = ip->root; | |
218 | struct btrfs_trans_handle *trans; | |
219 | unsigned int flags, oldflags; | |
220 | int ret; | |
f062abf0 LZ |
221 | u64 ip_oldflags; |
222 | unsigned int i_oldflags; | |
7e97b8da | 223 | umode_t mode; |
6cbff00f | 224 | |
bd60ea0f DS |
225 | if (!inode_owner_or_capable(inode)) |
226 | return -EPERM; | |
227 | ||
b83cc969 LZ |
228 | if (btrfs_root_readonly(root)) |
229 | return -EROFS; | |
230 | ||
6cbff00f CH |
231 | if (copy_from_user(&flags, arg, sizeof(flags))) |
232 | return -EFAULT; | |
233 | ||
75e7cb7f LB |
234 | ret = check_flags(flags); |
235 | if (ret) | |
236 | return ret; | |
f46b5a66 | 237 | |
e7848683 JK |
238 | ret = mnt_want_write_file(file); |
239 | if (ret) | |
240 | return ret; | |
241 | ||
6cbff00f CH |
242 | mutex_lock(&inode->i_mutex); |
243 | ||
f062abf0 LZ |
244 | ip_oldflags = ip->flags; |
245 | i_oldflags = inode->i_flags; | |
7e97b8da | 246 | mode = inode->i_mode; |
f062abf0 | 247 | |
6cbff00f CH |
248 | flags = btrfs_mask_flags(inode->i_mode, flags); |
249 | oldflags = btrfs_flags_to_ioctl(ip->flags); | |
250 | if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { | |
251 | if (!capable(CAP_LINUX_IMMUTABLE)) { | |
252 | ret = -EPERM; | |
253 | goto out_unlock; | |
254 | } | |
255 | } | |
256 | ||
6cbff00f CH |
257 | if (flags & FS_SYNC_FL) |
258 | ip->flags |= BTRFS_INODE_SYNC; | |
259 | else | |
260 | ip->flags &= ~BTRFS_INODE_SYNC; | |
261 | if (flags & FS_IMMUTABLE_FL) | |
262 | ip->flags |= BTRFS_INODE_IMMUTABLE; | |
263 | else | |
264 | ip->flags &= ~BTRFS_INODE_IMMUTABLE; | |
265 | if (flags & FS_APPEND_FL) | |
266 | ip->flags |= BTRFS_INODE_APPEND; | |
267 | else | |
268 | ip->flags &= ~BTRFS_INODE_APPEND; | |
269 | if (flags & FS_NODUMP_FL) | |
270 | ip->flags |= BTRFS_INODE_NODUMP; | |
271 | else | |
272 | ip->flags &= ~BTRFS_INODE_NODUMP; | |
273 | if (flags & FS_NOATIME_FL) | |
274 | ip->flags |= BTRFS_INODE_NOATIME; | |
275 | else | |
276 | ip->flags &= ~BTRFS_INODE_NOATIME; | |
277 | if (flags & FS_DIRSYNC_FL) | |
278 | ip->flags |= BTRFS_INODE_DIRSYNC; | |
279 | else | |
280 | ip->flags &= ~BTRFS_INODE_DIRSYNC; | |
7e97b8da DS |
281 | if (flags & FS_NOCOW_FL) { |
282 | if (S_ISREG(mode)) { | |
283 | /* | |
284 | * It's safe to turn csums off here, no extents exist. | |
285 | * Otherwise we want the flag to reflect the real COW | |
286 | * status of the file and will not set it. | |
287 | */ | |
288 | if (inode->i_size == 0) | |
289 | ip->flags |= BTRFS_INODE_NODATACOW | |
290 | | BTRFS_INODE_NODATASUM; | |
291 | } else { | |
292 | ip->flags |= BTRFS_INODE_NODATACOW; | |
293 | } | |
294 | } else { | |
295 | /* | |
296 | * Revert back under same assuptions as above | |
297 | */ | |
298 | if (S_ISREG(mode)) { | |
299 | if (inode->i_size == 0) | |
300 | ip->flags &= ~(BTRFS_INODE_NODATACOW | |
301 | | BTRFS_INODE_NODATASUM); | |
302 | } else { | |
303 | ip->flags &= ~BTRFS_INODE_NODATACOW; | |
304 | } | |
305 | } | |
6cbff00f | 306 | |
75e7cb7f LB |
307 | /* |
308 | * The COMPRESS flag can only be changed by users, while the NOCOMPRESS | |
309 | * flag may be changed automatically if compression code won't make | |
310 | * things smaller. | |
311 | */ | |
312 | if (flags & FS_NOCOMP_FL) { | |
313 | ip->flags &= ~BTRFS_INODE_COMPRESS; | |
314 | ip->flags |= BTRFS_INODE_NOCOMPRESS; | |
63541927 FDBM |
315 | |
316 | ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); | |
317 | if (ret && ret != -ENODATA) | |
318 | goto out_drop; | |
75e7cb7f | 319 | } else if (flags & FS_COMPR_FL) { |
63541927 FDBM |
320 | const char *comp; |
321 | ||
75e7cb7f LB |
322 | ip->flags |= BTRFS_INODE_COMPRESS; |
323 | ip->flags &= ~BTRFS_INODE_NOCOMPRESS; | |
63541927 FDBM |
324 | |
325 | if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO) | |
326 | comp = "lzo"; | |
327 | else | |
328 | comp = "zlib"; | |
329 | ret = btrfs_set_prop(inode, "btrfs.compression", | |
330 | comp, strlen(comp), 0); | |
331 | if (ret) | |
332 | goto out_drop; | |
333 | ||
ebcb904d | 334 | } else { |
78a017a2 FM |
335 | ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); |
336 | if (ret && ret != -ENODATA) | |
337 | goto out_drop; | |
ebcb904d | 338 | ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); |
75e7cb7f | 339 | } |
6cbff00f | 340 | |
4da6f1a3 | 341 | trans = btrfs_start_transaction(root, 1); |
f062abf0 LZ |
342 | if (IS_ERR(trans)) { |
343 | ret = PTR_ERR(trans); | |
344 | goto out_drop; | |
345 | } | |
6cbff00f | 346 | |
306424cc | 347 | btrfs_update_iflags(inode); |
0c4d2d95 | 348 | inode_inc_iversion(inode); |
306424cc | 349 | inode->i_ctime = CURRENT_TIME; |
6cbff00f | 350 | ret = btrfs_update_inode(trans, root, inode); |
6cbff00f | 351 | |
6cbff00f | 352 | btrfs_end_transaction(trans, root); |
f062abf0 LZ |
353 | out_drop: |
354 | if (ret) { | |
355 | ip->flags = ip_oldflags; | |
356 | inode->i_flags = i_oldflags; | |
357 | } | |
6cbff00f | 358 | |
6cbff00f CH |
359 | out_unlock: |
360 | mutex_unlock(&inode->i_mutex); | |
e7848683 | 361 | mnt_drop_write_file(file); |
2d4e6f6a | 362 | return ret; |
6cbff00f CH |
363 | } |
364 | ||
365 | static int btrfs_ioctl_getversion(struct file *file, int __user *arg) | |
366 | { | |
496ad9aa | 367 | struct inode *inode = file_inode(file); |
6cbff00f CH |
368 | |
369 | return put_user(inode->i_generation, arg); | |
370 | } | |
f46b5a66 | 371 | |
f7039b1d LD |
372 | static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) |
373 | { | |
54563d41 | 374 | struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); |
f7039b1d LD |
375 | struct btrfs_device *device; |
376 | struct request_queue *q; | |
377 | struct fstrim_range range; | |
378 | u64 minlen = ULLONG_MAX; | |
379 | u64 num_devices = 0; | |
815745cf | 380 | u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy); |
f7039b1d LD |
381 | int ret; |
382 | ||
383 | if (!capable(CAP_SYS_ADMIN)) | |
384 | return -EPERM; | |
385 | ||
1f78160c XG |
386 | rcu_read_lock(); |
387 | list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, | |
388 | dev_list) { | |
f7039b1d LD |
389 | if (!device->bdev) |
390 | continue; | |
391 | q = bdev_get_queue(device->bdev); | |
392 | if (blk_queue_discard(q)) { | |
393 | num_devices++; | |
394 | minlen = min((u64)q->limits.discard_granularity, | |
395 | minlen); | |
396 | } | |
397 | } | |
1f78160c | 398 | rcu_read_unlock(); |
f4c697e6 | 399 | |
f7039b1d LD |
400 | if (!num_devices) |
401 | return -EOPNOTSUPP; | |
f7039b1d LD |
402 | if (copy_from_user(&range, arg, sizeof(range))) |
403 | return -EFAULT; | |
e515c18b LC |
404 | if (range.start > total_bytes || |
405 | range.len < fs_info->sb->s_blocksize) | |
f4c697e6 | 406 | return -EINVAL; |
f7039b1d | 407 | |
f4c697e6 | 408 | range.len = min(range.len, total_bytes - range.start); |
f7039b1d | 409 | range.minlen = max(range.minlen, minlen); |
815745cf | 410 | ret = btrfs_trim_fs(fs_info->tree_root, &range); |
f7039b1d LD |
411 | if (ret < 0) |
412 | return ret; | |
413 | ||
414 | if (copy_to_user(arg, &range, sizeof(range))) | |
415 | return -EFAULT; | |
416 | ||
417 | return 0; | |
418 | } | |
419 | ||
dd5f9615 SB |
420 | int btrfs_is_empty_uuid(u8 *uuid) |
421 | { | |
46e0f66a CM |
422 | int i; |
423 | ||
424 | for (i = 0; i < BTRFS_UUID_SIZE; i++) { | |
425 | if (uuid[i]) | |
426 | return 0; | |
427 | } | |
428 | return 1; | |
dd5f9615 SB |
429 | } |
430 | ||
d5c12070 | 431 | static noinline int create_subvol(struct inode *dir, |
cb8e7090 | 432 | struct dentry *dentry, |
72fd032e | 433 | char *name, int namelen, |
6f72c7e2 | 434 | u64 *async_transid, |
8696c533 | 435 | struct btrfs_qgroup_inherit *inherit) |
f46b5a66 CH |
436 | { |
437 | struct btrfs_trans_handle *trans; | |
438 | struct btrfs_key key; | |
439 | struct btrfs_root_item root_item; | |
440 | struct btrfs_inode_item *inode_item; | |
441 | struct extent_buffer *leaf; | |
d5c12070 | 442 | struct btrfs_root *root = BTRFS_I(dir)->root; |
76dda93c | 443 | struct btrfs_root *new_root; |
d5c12070 | 444 | struct btrfs_block_rsv block_rsv; |
8ea05e3a | 445 | struct timespec cur_time = CURRENT_TIME; |
5662344b | 446 | struct inode *inode; |
f46b5a66 CH |
447 | int ret; |
448 | int err; | |
449 | u64 objectid; | |
450 | u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; | |
3de4586c | 451 | u64 index = 0; |
d5c12070 | 452 | u64 qgroup_reserved; |
8ea05e3a | 453 | uuid_le new_uuid; |
f46b5a66 | 454 | |
581bb050 | 455 | ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid); |
2fbe8c8a | 456 | if (ret) |
a22285a6 | 457 | return ret; |
6a912213 | 458 | |
e09fe2d2 QW |
459 | /* |
460 | * Don't create subvolume whose level is not zero. Or qgroup will be | |
461 | * screwed up since it assume subvolme qgroup's level to be 0. | |
462 | */ | |
463 | if (btrfs_qgroup_level(objectid)) | |
464 | return -ENOSPC; | |
465 | ||
d5c12070 | 466 | btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); |
9ed74f2d | 467 | /* |
d5c12070 MX |
468 | * The same as the snapshot creation, please see the comment |
469 | * of create_snapshot(). | |
9ed74f2d | 470 | */ |
d5c12070 | 471 | ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, |
dd5f9615 | 472 | 8, &qgroup_reserved, false); |
d5c12070 MX |
473 | if (ret) |
474 | return ret; | |
475 | ||
476 | trans = btrfs_start_transaction(root, 0); | |
477 | if (IS_ERR(trans)) { | |
478 | ret = PTR_ERR(trans); | |
de6e8200 LB |
479 | btrfs_subvolume_release_metadata(root, &block_rsv, |
480 | qgroup_reserved); | |
481 | return ret; | |
d5c12070 MX |
482 | } |
483 | trans->block_rsv = &block_rsv; | |
484 | trans->bytes_reserved = block_rsv.size; | |
f46b5a66 | 485 | |
8696c533 | 486 | ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit); |
6f72c7e2 AJ |
487 | if (ret) |
488 | goto fail; | |
489 | ||
4d75f8a9 | 490 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); |
8e8a1e31 JB |
491 | if (IS_ERR(leaf)) { |
492 | ret = PTR_ERR(leaf); | |
493 | goto fail; | |
494 | } | |
f46b5a66 | 495 | |
5d4f98a2 | 496 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
f46b5a66 CH |
497 | btrfs_set_header_bytenr(leaf, leaf->start); |
498 | btrfs_set_header_generation(leaf, trans->transid); | |
5d4f98a2 | 499 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); |
f46b5a66 CH |
500 | btrfs_set_header_owner(leaf, objectid); |
501 | ||
0a4e5586 | 502 | write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(), |
f46b5a66 | 503 | BTRFS_FSID_SIZE); |
5d4f98a2 | 504 | write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, |
b308bc2f | 505 | btrfs_header_chunk_tree_uuid(leaf), |
5d4f98a2 | 506 | BTRFS_UUID_SIZE); |
f46b5a66 CH |
507 | btrfs_mark_buffer_dirty(leaf); |
508 | ||
8ea05e3a AB |
509 | memset(&root_item, 0, sizeof(root_item)); |
510 | ||
f46b5a66 | 511 | inode_item = &root_item.inode; |
3cae210f QW |
512 | btrfs_set_stack_inode_generation(inode_item, 1); |
513 | btrfs_set_stack_inode_size(inode_item, 3); | |
514 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
707e8a07 | 515 | btrfs_set_stack_inode_nbytes(inode_item, root->nodesize); |
3cae210f | 516 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
f46b5a66 | 517 | |
3cae210f QW |
518 | btrfs_set_root_flags(&root_item, 0); |
519 | btrfs_set_root_limit(&root_item, 0); | |
520 | btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); | |
08fe4db1 | 521 | |
f46b5a66 | 522 | btrfs_set_root_bytenr(&root_item, leaf->start); |
84234f3a | 523 | btrfs_set_root_generation(&root_item, trans->transid); |
f46b5a66 CH |
524 | btrfs_set_root_level(&root_item, 0); |
525 | btrfs_set_root_refs(&root_item, 1); | |
86b9f2ec | 526 | btrfs_set_root_used(&root_item, leaf->len); |
80ff3856 | 527 | btrfs_set_root_last_snapshot(&root_item, 0); |
f46b5a66 | 528 | |
8ea05e3a AB |
529 | btrfs_set_root_generation_v2(&root_item, |
530 | btrfs_root_generation(&root_item)); | |
531 | uuid_le_gen(&new_uuid); | |
532 | memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE); | |
3cae210f QW |
533 | btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec); |
534 | btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec); | |
8ea05e3a AB |
535 | root_item.ctime = root_item.otime; |
536 | btrfs_set_root_ctransid(&root_item, trans->transid); | |
537 | btrfs_set_root_otransid(&root_item, trans->transid); | |
f46b5a66 | 538 | |
925baedd | 539 | btrfs_tree_unlock(leaf); |
f46b5a66 CH |
540 | free_extent_buffer(leaf); |
541 | leaf = NULL; | |
542 | ||
543 | btrfs_set_root_dirid(&root_item, new_dirid); | |
544 | ||
545 | key.objectid = objectid; | |
5d4f98a2 | 546 | key.offset = 0; |
962a298f | 547 | key.type = BTRFS_ROOT_ITEM_KEY; |
f46b5a66 CH |
548 | ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, |
549 | &root_item); | |
550 | if (ret) | |
551 | goto fail; | |
552 | ||
76dda93c YZ |
553 | key.offset = (u64)-1; |
554 | new_root = btrfs_read_fs_root_no_name(root->fs_info, &key); | |
79787eaa JM |
555 | if (IS_ERR(new_root)) { |
556 | btrfs_abort_transaction(trans, root, PTR_ERR(new_root)); | |
557 | ret = PTR_ERR(new_root); | |
558 | goto fail; | |
559 | } | |
76dda93c YZ |
560 | |
561 | btrfs_record_root_in_trans(trans, new_root); | |
562 | ||
63541927 | 563 | ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); |
ce598979 MF |
564 | if (ret) { |
565 | /* We potentially lose an unused inode item here */ | |
79787eaa | 566 | btrfs_abort_transaction(trans, root, ret); |
ce598979 MF |
567 | goto fail; |
568 | } | |
569 | ||
f46b5a66 CH |
570 | /* |
571 | * insert the directory item | |
572 | */ | |
3de4586c | 573 | ret = btrfs_set_inode_index(dir, &index); |
79787eaa JM |
574 | if (ret) { |
575 | btrfs_abort_transaction(trans, root, ret); | |
576 | goto fail; | |
577 | } | |
3de4586c CM |
578 | |
579 | ret = btrfs_insert_dir_item(trans, root, | |
16cdcec7 | 580 | name, namelen, dir, &key, |
3de4586c | 581 | BTRFS_FT_DIR, index); |
79787eaa JM |
582 | if (ret) { |
583 | btrfs_abort_transaction(trans, root, ret); | |
f46b5a66 | 584 | goto fail; |
79787eaa | 585 | } |
0660b5af | 586 | |
52c26179 YZ |
587 | btrfs_i_size_write(dir, dir->i_size + namelen * 2); |
588 | ret = btrfs_update_inode(trans, root, dir); | |
589 | BUG_ON(ret); | |
590 | ||
0660b5af | 591 | ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, |
4df27c4d | 592 | objectid, root->root_key.objectid, |
33345d01 | 593 | btrfs_ino(dir), index, name, namelen); |
76dda93c | 594 | BUG_ON(ret); |
f46b5a66 | 595 | |
dd5f9615 SB |
596 | ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root, |
597 | root_item.uuid, BTRFS_UUID_KEY_SUBVOL, | |
598 | objectid); | |
599 | if (ret) | |
600 | btrfs_abort_transaction(trans, root, ret); | |
601 | ||
f46b5a66 | 602 | fail: |
d5c12070 MX |
603 | trans->block_rsv = NULL; |
604 | trans->bytes_reserved = 0; | |
de6e8200 LB |
605 | btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); |
606 | ||
72fd032e SW |
607 | if (async_transid) { |
608 | *async_transid = trans->transid; | |
609 | err = btrfs_commit_transaction_async(trans, root, 1); | |
00d71c9c MX |
610 | if (err) |
611 | err = btrfs_commit_transaction(trans, root); | |
72fd032e SW |
612 | } else { |
613 | err = btrfs_commit_transaction(trans, root); | |
614 | } | |
f46b5a66 CH |
615 | if (err && !ret) |
616 | ret = err; | |
1a65e24b | 617 | |
5662344b TI |
618 | if (!ret) { |
619 | inode = btrfs_lookup_dentry(dir, dentry); | |
de6e8200 LB |
620 | if (IS_ERR(inode)) |
621 | return PTR_ERR(inode); | |
5662344b TI |
622 | d_instantiate(dentry, inode); |
623 | } | |
f46b5a66 CH |
624 | return ret; |
625 | } | |
626 | ||
9ea24bbe | 627 | static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root) |
8257b2dc MX |
628 | { |
629 | s64 writers; | |
630 | DEFINE_WAIT(wait); | |
631 | ||
632 | do { | |
633 | prepare_to_wait(&root->subv_writers->wait, &wait, | |
634 | TASK_UNINTERRUPTIBLE); | |
635 | ||
636 | writers = percpu_counter_sum(&root->subv_writers->counter); | |
637 | if (writers) | |
638 | schedule(); | |
639 | ||
640 | finish_wait(&root->subv_writers->wait, &wait); | |
641 | } while (writers); | |
642 | } | |
643 | ||
e9662f70 MX |
644 | static int create_snapshot(struct btrfs_root *root, struct inode *dir, |
645 | struct dentry *dentry, char *name, int namelen, | |
646 | u64 *async_transid, bool readonly, | |
647 | struct btrfs_qgroup_inherit *inherit) | |
f46b5a66 | 648 | { |
2e4bfab9 | 649 | struct inode *inode; |
f46b5a66 CH |
650 | struct btrfs_pending_snapshot *pending_snapshot; |
651 | struct btrfs_trans_handle *trans; | |
2e4bfab9 | 652 | int ret; |
f46b5a66 | 653 | |
27cdeb70 | 654 | if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) |
f46b5a66 CH |
655 | return -EINVAL; |
656 | ||
8257b2dc | 657 | atomic_inc(&root->will_be_snapshoted); |
4e857c58 | 658 | smp_mb__after_atomic(); |
9ea24bbe | 659 | btrfs_wait_for_no_snapshoting_writes(root); |
8257b2dc | 660 | |
6a03843d MX |
661 | ret = btrfs_start_delalloc_inodes(root, 0); |
662 | if (ret) | |
8257b2dc | 663 | goto out; |
6a03843d | 664 | |
b0244199 | 665 | btrfs_wait_ordered_extents(root, -1); |
6a03843d | 666 | |
3de4586c | 667 | pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); |
8257b2dc MX |
668 | if (!pending_snapshot) { |
669 | ret = -ENOMEM; | |
670 | goto out; | |
671 | } | |
a22285a6 | 672 | |
66d8f3dd MX |
673 | btrfs_init_block_rsv(&pending_snapshot->block_rsv, |
674 | BTRFS_BLOCK_RSV_TEMP); | |
d5c12070 MX |
675 | /* |
676 | * 1 - parent dir inode | |
677 | * 2 - dir entries | |
678 | * 1 - root item | |
679 | * 2 - root ref/backref | |
680 | * 1 - root of snapshot | |
dd5f9615 | 681 | * 1 - UUID item |
d5c12070 MX |
682 | */ |
683 | ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, | |
dd5f9615 | 684 | &pending_snapshot->block_rsv, 8, |
ee3441b4 JM |
685 | &pending_snapshot->qgroup_reserved, |
686 | false); | |
d5c12070 | 687 | if (ret) |
8257b2dc | 688 | goto free; |
d5c12070 | 689 | |
3de4586c | 690 | pending_snapshot->dentry = dentry; |
f46b5a66 | 691 | pending_snapshot->root = root; |
b83cc969 | 692 | pending_snapshot->readonly = readonly; |
e9662f70 | 693 | pending_snapshot->dir = dir; |
8696c533 | 694 | pending_snapshot->inherit = inherit; |
a22285a6 | 695 | |
d5c12070 | 696 | trans = btrfs_start_transaction(root, 0); |
a22285a6 YZ |
697 | if (IS_ERR(trans)) { |
698 | ret = PTR_ERR(trans); | |
699 | goto fail; | |
700 | } | |
701 | ||
8351583e | 702 | spin_lock(&root->fs_info->trans_lock); |
f46b5a66 CH |
703 | list_add(&pending_snapshot->list, |
704 | &trans->transaction->pending_snapshots); | |
8351583e | 705 | spin_unlock(&root->fs_info->trans_lock); |
72fd032e SW |
706 | if (async_transid) { |
707 | *async_transid = trans->transid; | |
708 | ret = btrfs_commit_transaction_async(trans, | |
709 | root->fs_info->extent_root, 1); | |
00d71c9c MX |
710 | if (ret) |
711 | ret = btrfs_commit_transaction(trans, root); | |
72fd032e SW |
712 | } else { |
713 | ret = btrfs_commit_transaction(trans, | |
714 | root->fs_info->extent_root); | |
715 | } | |
aec8030a | 716 | if (ret) |
c37b2b62 | 717 | goto fail; |
a22285a6 YZ |
718 | |
719 | ret = pending_snapshot->error; | |
720 | if (ret) | |
721 | goto fail; | |
722 | ||
d3797308 CM |
723 | ret = btrfs_orphan_cleanup(pending_snapshot->snap); |
724 | if (ret) | |
725 | goto fail; | |
726 | ||
2b0143b5 | 727 | inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); |
2e4bfab9 YZ |
728 | if (IS_ERR(inode)) { |
729 | ret = PTR_ERR(inode); | |
730 | goto fail; | |
731 | } | |
5662344b | 732 | |
2e4bfab9 YZ |
733 | d_instantiate(dentry, inode); |
734 | ret = 0; | |
735 | fail: | |
d5c12070 MX |
736 | btrfs_subvolume_release_metadata(BTRFS_I(dir)->root, |
737 | &pending_snapshot->block_rsv, | |
738 | pending_snapshot->qgroup_reserved); | |
8257b2dc | 739 | free: |
a22285a6 | 740 | kfree(pending_snapshot); |
8257b2dc | 741 | out: |
9ea24bbe FM |
742 | if (atomic_dec_and_test(&root->will_be_snapshoted)) |
743 | wake_up_atomic_t(&root->will_be_snapshoted); | |
f46b5a66 CH |
744 | return ret; |
745 | } | |
746 | ||
4260f7c7 SW |
747 | /* copy of may_delete in fs/namei.c() |
748 | * Check whether we can remove a link victim from directory dir, check | |
749 | * whether the type of victim is right. | |
750 | * 1. We can't do it if dir is read-only (done in permission()) | |
751 | * 2. We should have write and exec permissions on dir | |
752 | * 3. We can't remove anything from append-only dir | |
753 | * 4. We can't do anything with immutable dir (done in permission()) | |
754 | * 5. If the sticky bit on dir is set we should either | |
755 | * a. be owner of dir, or | |
756 | * b. be owner of victim, or | |
757 | * c. have CAP_FOWNER capability | |
758 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
759 | * links pointing to it. | |
760 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
761 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
762 | * 9. We can't remove a root or mountpoint. | |
763 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
764 | * nfs_async_unlink(). | |
765 | */ | |
766 | ||
67871254 | 767 | static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) |
4260f7c7 SW |
768 | { |
769 | int error; | |
770 | ||
2b0143b5 | 771 | if (d_really_is_negative(victim)) |
4260f7c7 SW |
772 | return -ENOENT; |
773 | ||
2b0143b5 | 774 | BUG_ON(d_inode(victim->d_parent) != dir); |
4fa6b5ec | 775 | audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); |
4260f7c7 SW |
776 | |
777 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); | |
778 | if (error) | |
779 | return error; | |
780 | if (IS_APPEND(dir)) | |
781 | return -EPERM; | |
2b0143b5 DH |
782 | if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || |
783 | IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) | |
4260f7c7 SW |
784 | return -EPERM; |
785 | if (isdir) { | |
e36cb0b8 | 786 | if (!d_is_dir(victim)) |
4260f7c7 SW |
787 | return -ENOTDIR; |
788 | if (IS_ROOT(victim)) | |
789 | return -EBUSY; | |
e36cb0b8 | 790 | } else if (d_is_dir(victim)) |
4260f7c7 SW |
791 | return -EISDIR; |
792 | if (IS_DEADDIR(dir)) | |
793 | return -ENOENT; | |
794 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
795 | return -EBUSY; | |
796 | return 0; | |
797 | } | |
798 | ||
cb8e7090 CH |
799 | /* copy of may_create in fs/namei.c() */ |
800 | static inline int btrfs_may_create(struct inode *dir, struct dentry *child) | |
801 | { | |
2b0143b5 | 802 | if (d_really_is_positive(child)) |
cb8e7090 CH |
803 | return -EEXIST; |
804 | if (IS_DEADDIR(dir)) | |
805 | return -ENOENT; | |
806 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); | |
807 | } | |
808 | ||
809 | /* | |
810 | * Create a new subvolume below @parent. This is largely modeled after | |
811 | * sys_mkdirat and vfs_mkdir, but we only do a single component lookup | |
812 | * inside this filesystem so it's quite a bit simpler. | |
813 | */ | |
76dda93c YZ |
814 | static noinline int btrfs_mksubvol(struct path *parent, |
815 | char *name, int namelen, | |
72fd032e | 816 | struct btrfs_root *snap_src, |
6f72c7e2 | 817 | u64 *async_transid, bool readonly, |
8696c533 | 818 | struct btrfs_qgroup_inherit *inherit) |
cb8e7090 | 819 | { |
2b0143b5 | 820 | struct inode *dir = d_inode(parent->dentry); |
cb8e7090 CH |
821 | struct dentry *dentry; |
822 | int error; | |
823 | ||
5c50c9b8 DS |
824 | error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT); |
825 | if (error == -EINTR) | |
826 | return error; | |
cb8e7090 CH |
827 | |
828 | dentry = lookup_one_len(name, parent->dentry, namelen); | |
829 | error = PTR_ERR(dentry); | |
830 | if (IS_ERR(dentry)) | |
831 | goto out_unlock; | |
832 | ||
833 | error = -EEXIST; | |
2b0143b5 | 834 | if (d_really_is_positive(dentry)) |
cb8e7090 CH |
835 | goto out_dput; |
836 | ||
76dda93c | 837 | error = btrfs_may_create(dir, dentry); |
cb8e7090 | 838 | if (error) |
a874a63e | 839 | goto out_dput; |
cb8e7090 | 840 | |
9c52057c CM |
841 | /* |
842 | * even if this name doesn't exist, we may get hash collisions. | |
843 | * check for them now when we can safely fail | |
844 | */ | |
845 | error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, | |
846 | dir->i_ino, name, | |
847 | namelen); | |
848 | if (error) | |
849 | goto out_dput; | |
850 | ||
76dda93c YZ |
851 | down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); |
852 | ||
853 | if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) | |
854 | goto out_up_read; | |
855 | ||
3de4586c | 856 | if (snap_src) { |
e9662f70 | 857 | error = create_snapshot(snap_src, dir, dentry, name, namelen, |
6f72c7e2 | 858 | async_transid, readonly, inherit); |
3de4586c | 859 | } else { |
d5c12070 MX |
860 | error = create_subvol(dir, dentry, name, namelen, |
861 | async_transid, inherit); | |
3de4586c | 862 | } |
76dda93c YZ |
863 | if (!error) |
864 | fsnotify_mkdir(dir, dentry); | |
865 | out_up_read: | |
866 | up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); | |
cb8e7090 CH |
867 | out_dput: |
868 | dput(dentry); | |
869 | out_unlock: | |
76dda93c | 870 | mutex_unlock(&dir->i_mutex); |
cb8e7090 CH |
871 | return error; |
872 | } | |
873 | ||
4cb5300b CM |
874 | /* |
875 | * When we're defragging a range, we don't want to kick it off again | |
876 | * if it is really just waiting for delalloc to send it down. | |
877 | * If we find a nice big extent or delalloc range for the bytes in the | |
878 | * file you want to defrag, we return 0 to let you know to skip this | |
879 | * part of the file | |
880 | */ | |
aab110ab | 881 | static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) |
4cb5300b CM |
882 | { |
883 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | |
884 | struct extent_map *em = NULL; | |
885 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
886 | u64 end; | |
887 | ||
888 | read_lock(&em_tree->lock); | |
889 | em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); | |
890 | read_unlock(&em_tree->lock); | |
891 | ||
892 | if (em) { | |
893 | end = extent_map_end(em); | |
894 | free_extent_map(em); | |
895 | if (end - offset > thresh) | |
896 | return 0; | |
897 | } | |
898 | /* if we already have a nice delalloc here, just stop */ | |
899 | thresh /= 2; | |
900 | end = count_range_bits(io_tree, &offset, offset + thresh, | |
901 | thresh, EXTENT_DELALLOC, 1); | |
902 | if (end >= thresh) | |
903 | return 0; | |
904 | return 1; | |
905 | } | |
906 | ||
907 | /* | |
908 | * helper function to walk through a file and find extents | |
909 | * newer than a specific transid, and smaller than thresh. | |
910 | * | |
911 | * This is used by the defragging code to find new and small | |
912 | * extents | |
913 | */ | |
914 | static int find_new_extents(struct btrfs_root *root, | |
915 | struct inode *inode, u64 newer_than, | |
aab110ab | 916 | u64 *off, u32 thresh) |
4cb5300b CM |
917 | { |
918 | struct btrfs_path *path; | |
919 | struct btrfs_key min_key; | |
4cb5300b CM |
920 | struct extent_buffer *leaf; |
921 | struct btrfs_file_extent_item *extent; | |
922 | int type; | |
923 | int ret; | |
a4689d2b | 924 | u64 ino = btrfs_ino(inode); |
4cb5300b CM |
925 | |
926 | path = btrfs_alloc_path(); | |
927 | if (!path) | |
928 | return -ENOMEM; | |
929 | ||
a4689d2b | 930 | min_key.objectid = ino; |
4cb5300b CM |
931 | min_key.type = BTRFS_EXTENT_DATA_KEY; |
932 | min_key.offset = *off; | |
933 | ||
67871254 | 934 | while (1) { |
6174d3cb | 935 | ret = btrfs_search_forward(root, &min_key, path, newer_than); |
4cb5300b CM |
936 | if (ret != 0) |
937 | goto none; | |
f094c9bd | 938 | process_slot: |
a4689d2b | 939 | if (min_key.objectid != ino) |
4cb5300b CM |
940 | goto none; |
941 | if (min_key.type != BTRFS_EXTENT_DATA_KEY) | |
942 | goto none; | |
943 | ||
944 | leaf = path->nodes[0]; | |
945 | extent = btrfs_item_ptr(leaf, path->slots[0], | |
946 | struct btrfs_file_extent_item); | |
947 | ||
948 | type = btrfs_file_extent_type(leaf, extent); | |
949 | if (type == BTRFS_FILE_EXTENT_REG && | |
950 | btrfs_file_extent_num_bytes(leaf, extent) < thresh && | |
951 | check_defrag_in_cache(inode, min_key.offset, thresh)) { | |
952 | *off = min_key.offset; | |
953 | btrfs_free_path(path); | |
954 | return 0; | |
955 | } | |
956 | ||
f094c9bd FM |
957 | path->slots[0]++; |
958 | if (path->slots[0] < btrfs_header_nritems(leaf)) { | |
959 | btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); | |
960 | goto process_slot; | |
961 | } | |
962 | ||
4cb5300b CM |
963 | if (min_key.offset == (u64)-1) |
964 | goto none; | |
965 | ||
966 | min_key.offset++; | |
967 | btrfs_release_path(path); | |
968 | } | |
969 | none: | |
970 | btrfs_free_path(path); | |
971 | return -ENOENT; | |
972 | } | |
973 | ||
6c282eb4 | 974 | static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) |
17ce6ef8 LB |
975 | { |
976 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
6c282eb4 LZ |
977 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
978 | struct extent_map *em; | |
979 | u64 len = PAGE_CACHE_SIZE; | |
17ce6ef8 | 980 | |
6c282eb4 LZ |
981 | /* |
982 | * hopefully we have this extent in the tree already, try without | |
983 | * the full extent lock | |
984 | */ | |
17ce6ef8 | 985 | read_lock(&em_tree->lock); |
6c282eb4 | 986 | em = lookup_extent_mapping(em_tree, start, len); |
17ce6ef8 LB |
987 | read_unlock(&em_tree->lock); |
988 | ||
6c282eb4 | 989 | if (!em) { |
308d9800 FM |
990 | struct extent_state *cached = NULL; |
991 | u64 end = start + len - 1; | |
992 | ||
6c282eb4 | 993 | /* get the big lock and read metadata off disk */ |
308d9800 | 994 | lock_extent_bits(io_tree, start, end, 0, &cached); |
6c282eb4 | 995 | em = btrfs_get_extent(inode, NULL, 0, start, len, 0); |
308d9800 | 996 | unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS); |
6c282eb4 LZ |
997 | |
998 | if (IS_ERR(em)) | |
999 | return NULL; | |
1000 | } | |
1001 | ||
1002 | return em; | |
1003 | } | |
17ce6ef8 | 1004 | |
6c282eb4 LZ |
1005 | static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) |
1006 | { | |
1007 | struct extent_map *next; | |
1008 | bool ret = true; | |
1009 | ||
1010 | /* this is the last extent */ | |
1011 | if (em->start + em->len >= i_size_read(inode)) | |
1012 | return false; | |
1013 | ||
1014 | next = defrag_lookup_extent(inode, em->start + em->len); | |
e9512d72 CM |
1015 | if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) |
1016 | ret = false; | |
1017 | else if ((em->block_start + em->block_len == next->block_start) && | |
1018 | (em->block_len > 128 * 1024 && next->block_len > 128 * 1024)) | |
6c282eb4 LZ |
1019 | ret = false; |
1020 | ||
1021 | free_extent_map(next); | |
17ce6ef8 LB |
1022 | return ret; |
1023 | } | |
1024 | ||
aab110ab | 1025 | static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, |
a43a2111 AM |
1026 | u64 *last_len, u64 *skip, u64 *defrag_end, |
1027 | int compress) | |
940100a4 | 1028 | { |
6c282eb4 | 1029 | struct extent_map *em; |
940100a4 | 1030 | int ret = 1; |
6c282eb4 | 1031 | bool next_mergeable = true; |
940100a4 CM |
1032 | |
1033 | /* | |
008873ea | 1034 | * make sure that once we start defragging an extent, we keep on |
940100a4 CM |
1035 | * defragging it |
1036 | */ | |
1037 | if (start < *defrag_end) | |
1038 | return 1; | |
1039 | ||
1040 | *skip = 0; | |
1041 | ||
6c282eb4 LZ |
1042 | em = defrag_lookup_extent(inode, start); |
1043 | if (!em) | |
1044 | return 0; | |
940100a4 CM |
1045 | |
1046 | /* this will cover holes, and inline extents */ | |
17ce6ef8 | 1047 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) { |
940100a4 | 1048 | ret = 0; |
17ce6ef8 LB |
1049 | goto out; |
1050 | } | |
1051 | ||
6c282eb4 | 1052 | next_mergeable = defrag_check_next_extent(inode, em); |
940100a4 | 1053 | /* |
6c282eb4 LZ |
1054 | * we hit a real extent, if it is big or the next extent is not a |
1055 | * real extent, don't bother defragging it | |
940100a4 | 1056 | */ |
a43a2111 | 1057 | if (!compress && (*last_len == 0 || *last_len >= thresh) && |
6c282eb4 | 1058 | (em->len >= thresh || !next_mergeable)) |
940100a4 | 1059 | ret = 0; |
17ce6ef8 | 1060 | out: |
940100a4 CM |
1061 | /* |
1062 | * last_len ends up being a counter of how many bytes we've defragged. | |
1063 | * every time we choose not to defrag an extent, we reset *last_len | |
1064 | * so that the next tiny extent will force a defrag. | |
1065 | * | |
1066 | * The end result of this is that tiny extents before a single big | |
1067 | * extent will force at least part of that big extent to be defragged. | |
1068 | */ | |
1069 | if (ret) { | |
940100a4 CM |
1070 | *defrag_end = extent_map_end(em); |
1071 | } else { | |
1072 | *last_len = 0; | |
1073 | *skip = extent_map_end(em); | |
1074 | *defrag_end = 0; | |
1075 | } | |
1076 | ||
1077 | free_extent_map(em); | |
1078 | return ret; | |
1079 | } | |
1080 | ||
4cb5300b CM |
1081 | /* |
1082 | * it doesn't do much good to defrag one or two pages | |
1083 | * at a time. This pulls in a nice chunk of pages | |
1084 | * to COW and defrag. | |
1085 | * | |
1086 | * It also makes sure the delalloc code has enough | |
1087 | * dirty data to avoid making new small extents as part | |
1088 | * of the defrag | |
1089 | * | |
1090 | * It's a good idea to start RA on this range | |
1091 | * before calling this. | |
1092 | */ | |
1093 | static int cluster_pages_for_defrag(struct inode *inode, | |
1094 | struct page **pages, | |
1095 | unsigned long start_index, | |
c41570c9 | 1096 | unsigned long num_pages) |
f46b5a66 | 1097 | { |
4cb5300b CM |
1098 | unsigned long file_end; |
1099 | u64 isize = i_size_read(inode); | |
1100 | u64 page_start; | |
1101 | u64 page_end; | |
1f12bd06 | 1102 | u64 page_cnt; |
4cb5300b CM |
1103 | int ret; |
1104 | int i; | |
1105 | int i_done; | |
3eaa2885 | 1106 | struct btrfs_ordered_extent *ordered; |
4cb5300b | 1107 | struct extent_state *cached_state = NULL; |
600a45e1 | 1108 | struct extent_io_tree *tree; |
3b16a4e3 | 1109 | gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); |
4cb5300b | 1110 | |
4cb5300b | 1111 | file_end = (isize - 1) >> PAGE_CACHE_SHIFT; |
1f12bd06 LB |
1112 | if (!isize || start_index > file_end) |
1113 | return 0; | |
1114 | ||
1115 | page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); | |
4cb5300b CM |
1116 | |
1117 | ret = btrfs_delalloc_reserve_space(inode, | |
1f12bd06 | 1118 | page_cnt << PAGE_CACHE_SHIFT); |
4cb5300b CM |
1119 | if (ret) |
1120 | return ret; | |
4cb5300b | 1121 | i_done = 0; |
600a45e1 | 1122 | tree = &BTRFS_I(inode)->io_tree; |
4cb5300b CM |
1123 | |
1124 | /* step one, lock all the pages */ | |
1f12bd06 | 1125 | for (i = 0; i < page_cnt; i++) { |
4cb5300b | 1126 | struct page *page; |
600a45e1 | 1127 | again: |
a94733d0 | 1128 | page = find_or_create_page(inode->i_mapping, |
600a45e1 | 1129 | start_index + i, mask); |
4cb5300b CM |
1130 | if (!page) |
1131 | break; | |
1132 | ||
600a45e1 MX |
1133 | page_start = page_offset(page); |
1134 | page_end = page_start + PAGE_CACHE_SIZE - 1; | |
1135 | while (1) { | |
308d9800 FM |
1136 | lock_extent_bits(tree, page_start, page_end, |
1137 | 0, &cached_state); | |
600a45e1 MX |
1138 | ordered = btrfs_lookup_ordered_extent(inode, |
1139 | page_start); | |
308d9800 FM |
1140 | unlock_extent_cached(tree, page_start, page_end, |
1141 | &cached_state, GFP_NOFS); | |
600a45e1 MX |
1142 | if (!ordered) |
1143 | break; | |
1144 | ||
1145 | unlock_page(page); | |
1146 | btrfs_start_ordered_extent(inode, ordered, 1); | |
1147 | btrfs_put_ordered_extent(ordered); | |
1148 | lock_page(page); | |
1f12bd06 LB |
1149 | /* |
1150 | * we unlocked the page above, so we need check if | |
1151 | * it was released or not. | |
1152 | */ | |
1153 | if (page->mapping != inode->i_mapping) { | |
1154 | unlock_page(page); | |
1155 | page_cache_release(page); | |
1156 | goto again; | |
1157 | } | |
600a45e1 MX |
1158 | } |
1159 | ||
4cb5300b CM |
1160 | if (!PageUptodate(page)) { |
1161 | btrfs_readpage(NULL, page); | |
1162 | lock_page(page); | |
1163 | if (!PageUptodate(page)) { | |
1164 | unlock_page(page); | |
1165 | page_cache_release(page); | |
1166 | ret = -EIO; | |
1167 | break; | |
1168 | } | |
1169 | } | |
600a45e1 | 1170 | |
600a45e1 MX |
1171 | if (page->mapping != inode->i_mapping) { |
1172 | unlock_page(page); | |
1173 | page_cache_release(page); | |
1174 | goto again; | |
1175 | } | |
1176 | ||
4cb5300b CM |
1177 | pages[i] = page; |
1178 | i_done++; | |
1179 | } | |
1180 | if (!i_done || ret) | |
1181 | goto out; | |
1182 | ||
1183 | if (!(inode->i_sb->s_flags & MS_ACTIVE)) | |
1184 | goto out; | |
1185 | ||
1186 | /* | |
1187 | * so now we have a nice long stream of locked | |
1188 | * and up to date pages, lets wait on them | |
1189 | */ | |
1190 | for (i = 0; i < i_done; i++) | |
1191 | wait_on_page_writeback(pages[i]); | |
1192 | ||
1193 | page_start = page_offset(pages[0]); | |
1194 | page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE; | |
1195 | ||
1196 | lock_extent_bits(&BTRFS_I(inode)->io_tree, | |
d0082371 | 1197 | page_start, page_end - 1, 0, &cached_state); |
4cb5300b CM |
1198 | clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, |
1199 | page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | | |
9e8a4a8b LB |
1200 | EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, |
1201 | &cached_state, GFP_NOFS); | |
4cb5300b | 1202 | |
1f12bd06 | 1203 | if (i_done != page_cnt) { |
9e0baf60 JB |
1204 | spin_lock(&BTRFS_I(inode)->lock); |
1205 | BTRFS_I(inode)->outstanding_extents++; | |
1206 | spin_unlock(&BTRFS_I(inode)->lock); | |
4cb5300b | 1207 | btrfs_delalloc_release_space(inode, |
1f12bd06 | 1208 | (page_cnt - i_done) << PAGE_CACHE_SHIFT); |
4cb5300b CM |
1209 | } |
1210 | ||
1211 | ||
9e8a4a8b LB |
1212 | set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, |
1213 | &cached_state, GFP_NOFS); | |
4cb5300b CM |
1214 | |
1215 | unlock_extent_cached(&BTRFS_I(inode)->io_tree, | |
1216 | page_start, page_end - 1, &cached_state, | |
1217 | GFP_NOFS); | |
1218 | ||
1219 | for (i = 0; i < i_done; i++) { | |
1220 | clear_page_dirty_for_io(pages[i]); | |
1221 | ClearPageChecked(pages[i]); | |
1222 | set_page_extent_mapped(pages[i]); | |
1223 | set_page_dirty(pages[i]); | |
1224 | unlock_page(pages[i]); | |
1225 | page_cache_release(pages[i]); | |
1226 | } | |
1227 | return i_done; | |
1228 | out: | |
1229 | for (i = 0; i < i_done; i++) { | |
1230 | unlock_page(pages[i]); | |
1231 | page_cache_release(pages[i]); | |
1232 | } | |
1f12bd06 | 1233 | btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT); |
4cb5300b CM |
1234 | return ret; |
1235 | ||
1236 | } | |
1237 | ||
1238 | int btrfs_defrag_file(struct inode *inode, struct file *file, | |
1239 | struct btrfs_ioctl_defrag_range_args *range, | |
1240 | u64 newer_than, unsigned long max_to_defrag) | |
1241 | { | |
1242 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
4cb5300b | 1243 | struct file_ra_state *ra = NULL; |
f46b5a66 | 1244 | unsigned long last_index; |
151a31b2 | 1245 | u64 isize = i_size_read(inode); |
940100a4 CM |
1246 | u64 last_len = 0; |
1247 | u64 skip = 0; | |
1248 | u64 defrag_end = 0; | |
4cb5300b | 1249 | u64 newer_off = range->start; |
f46b5a66 | 1250 | unsigned long i; |
008873ea | 1251 | unsigned long ra_index = 0; |
f46b5a66 | 1252 | int ret; |
4cb5300b | 1253 | int defrag_count = 0; |
1a419d85 | 1254 | int compress_type = BTRFS_COMPRESS_ZLIB; |
aab110ab | 1255 | u32 extent_thresh = range->extent_thresh; |
c41570c9 JM |
1256 | unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT; |
1257 | unsigned long cluster = max_cluster; | |
4cb5300b CM |
1258 | u64 new_align = ~((u64)128 * 1024 - 1); |
1259 | struct page **pages = NULL; | |
1260 | ||
0abd5b17 LB |
1261 | if (isize == 0) |
1262 | return 0; | |
1263 | ||
1264 | if (range->start >= isize) | |
1265 | return -EINVAL; | |
1a419d85 LZ |
1266 | |
1267 | if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { | |
1268 | if (range->compress_type > BTRFS_COMPRESS_TYPES) | |
1269 | return -EINVAL; | |
1270 | if (range->compress_type) | |
1271 | compress_type = range->compress_type; | |
1272 | } | |
f46b5a66 | 1273 | |
0abd5b17 LB |
1274 | if (extent_thresh == 0) |
1275 | extent_thresh = 256 * 1024; | |
940100a4 | 1276 | |
4cb5300b CM |
1277 | /* |
1278 | * if we were not given a file, allocate a readahead | |
1279 | * context | |
1280 | */ | |
1281 | if (!file) { | |
1282 | ra = kzalloc(sizeof(*ra), GFP_NOFS); | |
1283 | if (!ra) | |
1284 | return -ENOMEM; | |
1285 | file_ra_state_init(ra, inode->i_mapping); | |
1286 | } else { | |
1287 | ra = &file->f_ra; | |
1288 | } | |
1289 | ||
d9b0d9ba | 1290 | pages = kmalloc_array(max_cluster, sizeof(struct page *), |
4cb5300b CM |
1291 | GFP_NOFS); |
1292 | if (!pages) { | |
1293 | ret = -ENOMEM; | |
1294 | goto out_ra; | |
1295 | } | |
1296 | ||
1297 | /* find the last page to defrag */ | |
1e701a32 | 1298 | if (range->start + range->len > range->start) { |
151a31b2 | 1299 | last_index = min_t(u64, isize - 1, |
1e701a32 CM |
1300 | range->start + range->len - 1) >> PAGE_CACHE_SHIFT; |
1301 | } else { | |
151a31b2 | 1302 | last_index = (isize - 1) >> PAGE_CACHE_SHIFT; |
1e701a32 CM |
1303 | } |
1304 | ||
4cb5300b CM |
1305 | if (newer_than) { |
1306 | ret = find_new_extents(root, inode, newer_than, | |
1307 | &newer_off, 64 * 1024); | |
1308 | if (!ret) { | |
1309 | range->start = newer_off; | |
1310 | /* | |
1311 | * we always align our defrag to help keep | |
1312 | * the extents in the file evenly spaced | |
1313 | */ | |
1314 | i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; | |
4cb5300b CM |
1315 | } else |
1316 | goto out_ra; | |
1317 | } else { | |
1318 | i = range->start >> PAGE_CACHE_SHIFT; | |
1319 | } | |
1320 | if (!max_to_defrag) | |
7ec31b54 | 1321 | max_to_defrag = last_index + 1; |
4cb5300b | 1322 | |
2a0f7f57 LZ |
1323 | /* |
1324 | * make writeback starts from i, so the defrag range can be | |
1325 | * written sequentially. | |
1326 | */ | |
1327 | if (i < inode->i_mapping->writeback_index) | |
1328 | inode->i_mapping->writeback_index = i; | |
1329 | ||
f7f43cc8 | 1330 | while (i <= last_index && defrag_count < max_to_defrag && |
ed6078f7 | 1331 | (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) { |
4cb5300b CM |
1332 | /* |
1333 | * make sure we stop running if someone unmounts | |
1334 | * the FS | |
1335 | */ | |
1336 | if (!(inode->i_sb->s_flags & MS_ACTIVE)) | |
1337 | break; | |
1338 | ||
210549eb | 1339 | if (btrfs_defrag_cancelled(root->fs_info)) { |
efe120a0 | 1340 | printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n"); |
210549eb DS |
1341 | ret = -EAGAIN; |
1342 | break; | |
1343 | } | |
1344 | ||
66c26892 | 1345 | if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT, |
6c282eb4 | 1346 | extent_thresh, &last_len, &skip, |
a43a2111 AM |
1347 | &defrag_end, range->flags & |
1348 | BTRFS_DEFRAG_RANGE_COMPRESS)) { | |
940100a4 CM |
1349 | unsigned long next; |
1350 | /* | |
1351 | * the should_defrag function tells us how much to skip | |
1352 | * bump our counter by the suggested amount | |
1353 | */ | |
ed6078f7 | 1354 | next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE); |
940100a4 CM |
1355 | i = max(i + 1, next); |
1356 | continue; | |
1357 | } | |
008873ea LZ |
1358 | |
1359 | if (!newer_than) { | |
1360 | cluster = (PAGE_CACHE_ALIGN(defrag_end) >> | |
1361 | PAGE_CACHE_SHIFT) - i; | |
1362 | cluster = min(cluster, max_cluster); | |
1363 | } else { | |
1364 | cluster = max_cluster; | |
1365 | } | |
1366 | ||
008873ea LZ |
1367 | if (i + cluster > ra_index) { |
1368 | ra_index = max(i, ra_index); | |
1369 | btrfs_force_ra(inode->i_mapping, ra, file, ra_index, | |
1370 | cluster); | |
1371 | ra_index += max_cluster; | |
1372 | } | |
940100a4 | 1373 | |
ecb8bea8 | 1374 | mutex_lock(&inode->i_mutex); |
633085c7 FDBM |
1375 | if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) |
1376 | BTRFS_I(inode)->force_compress = compress_type; | |
008873ea | 1377 | ret = cluster_pages_for_defrag(inode, pages, i, cluster); |
ecb8bea8 LB |
1378 | if (ret < 0) { |
1379 | mutex_unlock(&inode->i_mutex); | |
4cb5300b | 1380 | goto out_ra; |
ecb8bea8 | 1381 | } |
4cb5300b CM |
1382 | |
1383 | defrag_count += ret; | |
d0e1d66b | 1384 | balance_dirty_pages_ratelimited(inode->i_mapping); |
ecb8bea8 | 1385 | mutex_unlock(&inode->i_mutex); |
4cb5300b CM |
1386 | |
1387 | if (newer_than) { | |
1388 | if (newer_off == (u64)-1) | |
1389 | break; | |
1390 | ||
e1f041e1 LB |
1391 | if (ret > 0) |
1392 | i += ret; | |
1393 | ||
4cb5300b CM |
1394 | newer_off = max(newer_off + 1, |
1395 | (u64)i << PAGE_CACHE_SHIFT); | |
1396 | ||
1397 | ret = find_new_extents(root, inode, | |
1398 | newer_than, &newer_off, | |
1399 | 64 * 1024); | |
1400 | if (!ret) { | |
1401 | range->start = newer_off; | |
1402 | i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; | |
4cb5300b CM |
1403 | } else { |
1404 | break; | |
f46b5a66 | 1405 | } |
4cb5300b | 1406 | } else { |
008873ea | 1407 | if (ret > 0) { |
cbcc8326 | 1408 | i += ret; |
008873ea LZ |
1409 | last_len += ret << PAGE_CACHE_SHIFT; |
1410 | } else { | |
cbcc8326 | 1411 | i++; |
008873ea LZ |
1412 | last_len = 0; |
1413 | } | |
f46b5a66 | 1414 | } |
f46b5a66 CH |
1415 | } |
1416 | ||
dec8ef90 | 1417 | if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { |
1e701a32 | 1418 | filemap_flush(inode->i_mapping); |
dec8ef90 FM |
1419 | if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, |
1420 | &BTRFS_I(inode)->runtime_flags)) | |
1421 | filemap_flush(inode->i_mapping); | |
1422 | } | |
1e701a32 CM |
1423 | |
1424 | if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { | |
1425 | /* the filemap_flush will queue IO into the worker threads, but | |
1426 | * we have to make sure the IO is actually started and that | |
1427 | * ordered extents get created before we return | |
1428 | */ | |
1429 | atomic_inc(&root->fs_info->async_submit_draining); | |
1430 | while (atomic_read(&root->fs_info->nr_async_submits) || | |
1431 | atomic_read(&root->fs_info->async_delalloc_pages)) { | |
1432 | wait_event(root->fs_info->async_submit_wait, | |
1433 | (atomic_read(&root->fs_info->nr_async_submits) == 0 && | |
1434 | atomic_read(&root->fs_info->async_delalloc_pages) == 0)); | |
1435 | } | |
1436 | atomic_dec(&root->fs_info->async_submit_draining); | |
1e701a32 CM |
1437 | } |
1438 | ||
1a419d85 | 1439 | if (range->compress_type == BTRFS_COMPRESS_LZO) { |
2b0ce2c2 | 1440 | btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO); |
1a419d85 LZ |
1441 | } |
1442 | ||
60ccf82f | 1443 | ret = defrag_count; |
940100a4 | 1444 | |
4cb5300b | 1445 | out_ra: |
633085c7 FDBM |
1446 | if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { |
1447 | mutex_lock(&inode->i_mutex); | |
1448 | BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE; | |
1449 | mutex_unlock(&inode->i_mutex); | |
1450 | } | |
4cb5300b CM |
1451 | if (!file) |
1452 | kfree(ra); | |
1453 | kfree(pages); | |
940100a4 | 1454 | return ret; |
f46b5a66 CH |
1455 | } |
1456 | ||
198605a8 | 1457 | static noinline int btrfs_ioctl_resize(struct file *file, |
76dda93c | 1458 | void __user *arg) |
f46b5a66 CH |
1459 | { |
1460 | u64 new_size; | |
1461 | u64 old_size; | |
1462 | u64 devid = 1; | |
496ad9aa | 1463 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
f46b5a66 CH |
1464 | struct btrfs_ioctl_vol_args *vol_args; |
1465 | struct btrfs_trans_handle *trans; | |
1466 | struct btrfs_device *device = NULL; | |
1467 | char *sizestr; | |
9a40f122 | 1468 | char *retptr; |
f46b5a66 CH |
1469 | char *devstr = NULL; |
1470 | int ret = 0; | |
f46b5a66 CH |
1471 | int mod = 0; |
1472 | ||
e441d54d CM |
1473 | if (!capable(CAP_SYS_ADMIN)) |
1474 | return -EPERM; | |
1475 | ||
198605a8 MX |
1476 | ret = mnt_want_write_file(file); |
1477 | if (ret) | |
1478 | return ret; | |
1479 | ||
5ac00add SB |
1480 | if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, |
1481 | 1)) { | |
97547676 | 1482 | mnt_drop_write_file(file); |
e57138b3 | 1483 | return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
c9e9f97b ID |
1484 | } |
1485 | ||
5ac00add | 1486 | mutex_lock(&root->fs_info->volume_mutex); |
dae7b665 | 1487 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
1488 | if (IS_ERR(vol_args)) { |
1489 | ret = PTR_ERR(vol_args); | |
1490 | goto out; | |
1491 | } | |
5516e595 MF |
1492 | |
1493 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
f46b5a66 | 1494 | |
f46b5a66 CH |
1495 | sizestr = vol_args->name; |
1496 | devstr = strchr(sizestr, ':'); | |
1497 | if (devstr) { | |
f46b5a66 CH |
1498 | sizestr = devstr + 1; |
1499 | *devstr = '\0'; | |
1500 | devstr = vol_args->name; | |
58dfae63 Z |
1501 | ret = kstrtoull(devstr, 10, &devid); |
1502 | if (ret) | |
1503 | goto out_free; | |
dfd79829 MX |
1504 | if (!devid) { |
1505 | ret = -EINVAL; | |
1506 | goto out_free; | |
1507 | } | |
efe120a0 | 1508 | btrfs_info(root->fs_info, "resizing devid %llu", devid); |
f46b5a66 | 1509 | } |
dba60f3f | 1510 | |
aa1b8cd4 | 1511 | device = btrfs_find_device(root->fs_info, devid, NULL, NULL); |
f46b5a66 | 1512 | if (!device) { |
efe120a0 | 1513 | btrfs_info(root->fs_info, "resizer unable to find device %llu", |
c1c9ff7c | 1514 | devid); |
dfd79829 | 1515 | ret = -ENODEV; |
c9e9f97b | 1516 | goto out_free; |
f46b5a66 | 1517 | } |
dba60f3f MX |
1518 | |
1519 | if (!device->writeable) { | |
efe120a0 FH |
1520 | btrfs_info(root->fs_info, |
1521 | "resizer unable to apply on readonly device %llu", | |
c1c9ff7c | 1522 | devid); |
dfd79829 | 1523 | ret = -EPERM; |
4e42ae1b LB |
1524 | goto out_free; |
1525 | } | |
1526 | ||
f46b5a66 CH |
1527 | if (!strcmp(sizestr, "max")) |
1528 | new_size = device->bdev->bd_inode->i_size; | |
1529 | else { | |
1530 | if (sizestr[0] == '-') { | |
1531 | mod = -1; | |
1532 | sizestr++; | |
1533 | } else if (sizestr[0] == '+') { | |
1534 | mod = 1; | |
1535 | sizestr++; | |
1536 | } | |
9a40f122 GH |
1537 | new_size = memparse(sizestr, &retptr); |
1538 | if (*retptr != '\0' || new_size == 0) { | |
f46b5a66 | 1539 | ret = -EINVAL; |
c9e9f97b | 1540 | goto out_free; |
f46b5a66 CH |
1541 | } |
1542 | } | |
1543 | ||
63a212ab | 1544 | if (device->is_tgtdev_for_dev_replace) { |
dfd79829 | 1545 | ret = -EPERM; |
63a212ab SB |
1546 | goto out_free; |
1547 | } | |
1548 | ||
7cc8e58d | 1549 | old_size = btrfs_device_get_total_bytes(device); |
f46b5a66 CH |
1550 | |
1551 | if (mod < 0) { | |
1552 | if (new_size > old_size) { | |
1553 | ret = -EINVAL; | |
c9e9f97b | 1554 | goto out_free; |
f46b5a66 CH |
1555 | } |
1556 | new_size = old_size - new_size; | |
1557 | } else if (mod > 0) { | |
eb8052e0 | 1558 | if (new_size > ULLONG_MAX - old_size) { |
902c68a4 | 1559 | ret = -ERANGE; |
eb8052e0 WF |
1560 | goto out_free; |
1561 | } | |
f46b5a66 CH |
1562 | new_size = old_size + new_size; |
1563 | } | |
1564 | ||
1565 | if (new_size < 256 * 1024 * 1024) { | |
1566 | ret = -EINVAL; | |
c9e9f97b | 1567 | goto out_free; |
f46b5a66 CH |
1568 | } |
1569 | if (new_size > device->bdev->bd_inode->i_size) { | |
1570 | ret = -EFBIG; | |
c9e9f97b | 1571 | goto out_free; |
f46b5a66 CH |
1572 | } |
1573 | ||
b8b93add | 1574 | new_size = div_u64(new_size, root->sectorsize); |
f46b5a66 CH |
1575 | new_size *= root->sectorsize; |
1576 | ||
efe120a0 | 1577 | printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n", |
c1c9ff7c | 1578 | rcu_str_deref(device->name), new_size); |
f46b5a66 CH |
1579 | |
1580 | if (new_size > old_size) { | |
a22285a6 | 1581 | trans = btrfs_start_transaction(root, 0); |
98d5dc13 TI |
1582 | if (IS_ERR(trans)) { |
1583 | ret = PTR_ERR(trans); | |
c9e9f97b | 1584 | goto out_free; |
98d5dc13 | 1585 | } |
f46b5a66 CH |
1586 | ret = btrfs_grow_device(trans, device, new_size); |
1587 | btrfs_commit_transaction(trans, root); | |
ece7d20e | 1588 | } else if (new_size < old_size) { |
f46b5a66 | 1589 | ret = btrfs_shrink_device(device, new_size); |
0253f40e | 1590 | } /* equal, nothing need to do */ |
f46b5a66 | 1591 | |
c9e9f97b | 1592 | out_free: |
f46b5a66 | 1593 | kfree(vol_args); |
c9e9f97b ID |
1594 | out: |
1595 | mutex_unlock(&root->fs_info->volume_mutex); | |
5ac00add | 1596 | atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); |
18f39c41 | 1597 | mnt_drop_write_file(file); |
f46b5a66 CH |
1598 | return ret; |
1599 | } | |
1600 | ||
72fd032e | 1601 | static noinline int btrfs_ioctl_snap_create_transid(struct file *file, |
6f72c7e2 AJ |
1602 | char *name, unsigned long fd, int subvol, |
1603 | u64 *transid, bool readonly, | |
8696c533 | 1604 | struct btrfs_qgroup_inherit *inherit) |
f46b5a66 | 1605 | { |
f46b5a66 | 1606 | int namelen; |
3de4586c | 1607 | int ret = 0; |
f46b5a66 | 1608 | |
a874a63e LB |
1609 | ret = mnt_want_write_file(file); |
1610 | if (ret) | |
1611 | goto out; | |
1612 | ||
72fd032e SW |
1613 | namelen = strlen(name); |
1614 | if (strchr(name, '/')) { | |
f46b5a66 | 1615 | ret = -EINVAL; |
a874a63e | 1616 | goto out_drop_write; |
f46b5a66 CH |
1617 | } |
1618 | ||
16780cab CM |
1619 | if (name[0] == '.' && |
1620 | (namelen == 1 || (name[1] == '.' && namelen == 2))) { | |
1621 | ret = -EEXIST; | |
a874a63e | 1622 | goto out_drop_write; |
16780cab CM |
1623 | } |
1624 | ||
3de4586c | 1625 | if (subvol) { |
72fd032e | 1626 | ret = btrfs_mksubvol(&file->f_path, name, namelen, |
6f72c7e2 | 1627 | NULL, transid, readonly, inherit); |
cb8e7090 | 1628 | } else { |
2903ff01 | 1629 | struct fd src = fdget(fd); |
3de4586c | 1630 | struct inode *src_inode; |
2903ff01 | 1631 | if (!src.file) { |
3de4586c | 1632 | ret = -EINVAL; |
a874a63e | 1633 | goto out_drop_write; |
3de4586c CM |
1634 | } |
1635 | ||
496ad9aa AV |
1636 | src_inode = file_inode(src.file); |
1637 | if (src_inode->i_sb != file_inode(file)->i_sb) { | |
efe120a0 FH |
1638 | btrfs_info(BTRFS_I(src_inode)->root->fs_info, |
1639 | "Snapshot src from another FS"); | |
23ad5b17 | 1640 | ret = -EXDEV; |
d0242061 DS |
1641 | } else if (!inode_owner_or_capable(src_inode)) { |
1642 | /* | |
1643 | * Subvolume creation is not restricted, but snapshots | |
1644 | * are limited to own subvolumes only | |
1645 | */ | |
1646 | ret = -EPERM; | |
ecd18815 AV |
1647 | } else { |
1648 | ret = btrfs_mksubvol(&file->f_path, name, namelen, | |
1649 | BTRFS_I(src_inode)->root, | |
1650 | transid, readonly, inherit); | |
3de4586c | 1651 | } |
2903ff01 | 1652 | fdput(src); |
cb8e7090 | 1653 | } |
a874a63e LB |
1654 | out_drop_write: |
1655 | mnt_drop_write_file(file); | |
f46b5a66 | 1656 | out: |
72fd032e SW |
1657 | return ret; |
1658 | } | |
1659 | ||
1660 | static noinline int btrfs_ioctl_snap_create(struct file *file, | |
fa0d2b9b | 1661 | void __user *arg, int subvol) |
72fd032e | 1662 | { |
fa0d2b9b | 1663 | struct btrfs_ioctl_vol_args *vol_args; |
72fd032e SW |
1664 | int ret; |
1665 | ||
fa0d2b9b LZ |
1666 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1667 | if (IS_ERR(vol_args)) | |
1668 | return PTR_ERR(vol_args); | |
1669 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
72fd032e | 1670 | |
fa0d2b9b | 1671 | ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, |
b83cc969 | 1672 | vol_args->fd, subvol, |
6f72c7e2 | 1673 | NULL, false, NULL); |
fdfb1e4f | 1674 | |
fa0d2b9b LZ |
1675 | kfree(vol_args); |
1676 | return ret; | |
1677 | } | |
fdfb1e4f | 1678 | |
fa0d2b9b LZ |
1679 | static noinline int btrfs_ioctl_snap_create_v2(struct file *file, |
1680 | void __user *arg, int subvol) | |
1681 | { | |
1682 | struct btrfs_ioctl_vol_args_v2 *vol_args; | |
1683 | int ret; | |
1684 | u64 transid = 0; | |
1685 | u64 *ptr = NULL; | |
b83cc969 | 1686 | bool readonly = false; |
6f72c7e2 | 1687 | struct btrfs_qgroup_inherit *inherit = NULL; |
75eaa0e2 | 1688 | |
fa0d2b9b LZ |
1689 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1690 | if (IS_ERR(vol_args)) | |
1691 | return PTR_ERR(vol_args); | |
1692 | vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; | |
75eaa0e2 | 1693 | |
b83cc969 | 1694 | if (vol_args->flags & |
6f72c7e2 AJ |
1695 | ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | |
1696 | BTRFS_SUBVOL_QGROUP_INHERIT)) { | |
b83cc969 | 1697 | ret = -EOPNOTSUPP; |
c47ca32d | 1698 | goto free_args; |
72fd032e | 1699 | } |
fa0d2b9b LZ |
1700 | |
1701 | if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) | |
1702 | ptr = &transid; | |
b83cc969 LZ |
1703 | if (vol_args->flags & BTRFS_SUBVOL_RDONLY) |
1704 | readonly = true; | |
6f72c7e2 AJ |
1705 | if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { |
1706 | if (vol_args->size > PAGE_CACHE_SIZE) { | |
1707 | ret = -EINVAL; | |
c47ca32d | 1708 | goto free_args; |
6f72c7e2 AJ |
1709 | } |
1710 | inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); | |
1711 | if (IS_ERR(inherit)) { | |
1712 | ret = PTR_ERR(inherit); | |
c47ca32d | 1713 | goto free_args; |
6f72c7e2 AJ |
1714 | } |
1715 | } | |
fa0d2b9b LZ |
1716 | |
1717 | ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, | |
6f72c7e2 | 1718 | vol_args->fd, subvol, ptr, |
8696c533 | 1719 | readonly, inherit); |
c47ca32d DC |
1720 | if (ret) |
1721 | goto free_inherit; | |
fa0d2b9b | 1722 | |
c47ca32d DC |
1723 | if (ptr && copy_to_user(arg + |
1724 | offsetof(struct btrfs_ioctl_vol_args_v2, | |
1725 | transid), | |
1726 | ptr, sizeof(*ptr))) | |
fa0d2b9b | 1727 | ret = -EFAULT; |
c47ca32d DC |
1728 | |
1729 | free_inherit: | |
6f72c7e2 | 1730 | kfree(inherit); |
c47ca32d DC |
1731 | free_args: |
1732 | kfree(vol_args); | |
f46b5a66 CH |
1733 | return ret; |
1734 | } | |
1735 | ||
0caa102d LZ |
1736 | static noinline int btrfs_ioctl_subvol_getflags(struct file *file, |
1737 | void __user *arg) | |
1738 | { | |
496ad9aa | 1739 | struct inode *inode = file_inode(file); |
0caa102d LZ |
1740 | struct btrfs_root *root = BTRFS_I(inode)->root; |
1741 | int ret = 0; | |
1742 | u64 flags = 0; | |
1743 | ||
33345d01 | 1744 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) |
0caa102d LZ |
1745 | return -EINVAL; |
1746 | ||
1747 | down_read(&root->fs_info->subvol_sem); | |
1748 | if (btrfs_root_readonly(root)) | |
1749 | flags |= BTRFS_SUBVOL_RDONLY; | |
1750 | up_read(&root->fs_info->subvol_sem); | |
1751 | ||
1752 | if (copy_to_user(arg, &flags, sizeof(flags))) | |
1753 | ret = -EFAULT; | |
1754 | ||
1755 | return ret; | |
1756 | } | |
1757 | ||
1758 | static noinline int btrfs_ioctl_subvol_setflags(struct file *file, | |
1759 | void __user *arg) | |
1760 | { | |
496ad9aa | 1761 | struct inode *inode = file_inode(file); |
0caa102d LZ |
1762 | struct btrfs_root *root = BTRFS_I(inode)->root; |
1763 | struct btrfs_trans_handle *trans; | |
1764 | u64 root_flags; | |
1765 | u64 flags; | |
1766 | int ret = 0; | |
1767 | ||
bd60ea0f DS |
1768 | if (!inode_owner_or_capable(inode)) |
1769 | return -EPERM; | |
1770 | ||
b9ca0664 LB |
1771 | ret = mnt_want_write_file(file); |
1772 | if (ret) | |
1773 | goto out; | |
0caa102d | 1774 | |
b9ca0664 LB |
1775 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { |
1776 | ret = -EINVAL; | |
1777 | goto out_drop_write; | |
1778 | } | |
0caa102d | 1779 | |
b9ca0664 LB |
1780 | if (copy_from_user(&flags, arg, sizeof(flags))) { |
1781 | ret = -EFAULT; | |
1782 | goto out_drop_write; | |
1783 | } | |
0caa102d | 1784 | |
b9ca0664 LB |
1785 | if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { |
1786 | ret = -EINVAL; | |
1787 | goto out_drop_write; | |
1788 | } | |
0caa102d | 1789 | |
b9ca0664 LB |
1790 | if (flags & ~BTRFS_SUBVOL_RDONLY) { |
1791 | ret = -EOPNOTSUPP; | |
1792 | goto out_drop_write; | |
1793 | } | |
0caa102d LZ |
1794 | |
1795 | down_write(&root->fs_info->subvol_sem); | |
1796 | ||
1797 | /* nothing to do */ | |
1798 | if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) | |
b9ca0664 | 1799 | goto out_drop_sem; |
0caa102d LZ |
1800 | |
1801 | root_flags = btrfs_root_flags(&root->root_item); | |
2c686537 | 1802 | if (flags & BTRFS_SUBVOL_RDONLY) { |
0caa102d LZ |
1803 | btrfs_set_root_flags(&root->root_item, |
1804 | root_flags | BTRFS_ROOT_SUBVOL_RDONLY); | |
2c686537 DS |
1805 | } else { |
1806 | /* | |
1807 | * Block RO -> RW transition if this subvolume is involved in | |
1808 | * send | |
1809 | */ | |
1810 | spin_lock(&root->root_item_lock); | |
1811 | if (root->send_in_progress == 0) { | |
1812 | btrfs_set_root_flags(&root->root_item, | |
0caa102d | 1813 | root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); |
2c686537 DS |
1814 | spin_unlock(&root->root_item_lock); |
1815 | } else { | |
1816 | spin_unlock(&root->root_item_lock); | |
1817 | btrfs_warn(root->fs_info, | |
1818 | "Attempt to set subvolume %llu read-write during send", | |
1819 | root->root_key.objectid); | |
1820 | ret = -EPERM; | |
1821 | goto out_drop_sem; | |
1822 | } | |
1823 | } | |
0caa102d LZ |
1824 | |
1825 | trans = btrfs_start_transaction(root, 1); | |
1826 | if (IS_ERR(trans)) { | |
1827 | ret = PTR_ERR(trans); | |
1828 | goto out_reset; | |
1829 | } | |
1830 | ||
b4dc2b8c | 1831 | ret = btrfs_update_root(trans, root->fs_info->tree_root, |
0caa102d LZ |
1832 | &root->root_key, &root->root_item); |
1833 | ||
1834 | btrfs_commit_transaction(trans, root); | |
1835 | out_reset: | |
1836 | if (ret) | |
1837 | btrfs_set_root_flags(&root->root_item, root_flags); | |
b9ca0664 | 1838 | out_drop_sem: |
0caa102d | 1839 | up_write(&root->fs_info->subvol_sem); |
b9ca0664 LB |
1840 | out_drop_write: |
1841 | mnt_drop_write_file(file); | |
1842 | out: | |
0caa102d LZ |
1843 | return ret; |
1844 | } | |
1845 | ||
76dda93c YZ |
1846 | /* |
1847 | * helper to check if the subvolume references other subvolumes | |
1848 | */ | |
1849 | static noinline int may_destroy_subvol(struct btrfs_root *root) | |
1850 | { | |
1851 | struct btrfs_path *path; | |
175a2b87 | 1852 | struct btrfs_dir_item *di; |
76dda93c | 1853 | struct btrfs_key key; |
175a2b87 | 1854 | u64 dir_id; |
76dda93c YZ |
1855 | int ret; |
1856 | ||
1857 | path = btrfs_alloc_path(); | |
1858 | if (!path) | |
1859 | return -ENOMEM; | |
1860 | ||
175a2b87 JB |
1861 | /* Make sure this root isn't set as the default subvol */ |
1862 | dir_id = btrfs_super_root_dir(root->fs_info->super_copy); | |
1863 | di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path, | |
1864 | dir_id, "default", 7, 0); | |
1865 | if (di && !IS_ERR(di)) { | |
1866 | btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key); | |
1867 | if (key.objectid == root->root_key.objectid) { | |
72de6b53 GS |
1868 | ret = -EPERM; |
1869 | btrfs_err(root->fs_info, "deleting default subvolume " | |
1870 | "%llu is not allowed", key.objectid); | |
175a2b87 JB |
1871 | goto out; |
1872 | } | |
1873 | btrfs_release_path(path); | |
1874 | } | |
1875 | ||
76dda93c YZ |
1876 | key.objectid = root->root_key.objectid; |
1877 | key.type = BTRFS_ROOT_REF_KEY; | |
1878 | key.offset = (u64)-1; | |
1879 | ||
1880 | ret = btrfs_search_slot(NULL, root->fs_info->tree_root, | |
1881 | &key, path, 0, 0); | |
1882 | if (ret < 0) | |
1883 | goto out; | |
1884 | BUG_ON(ret == 0); | |
1885 | ||
1886 | ret = 0; | |
1887 | if (path->slots[0] > 0) { | |
1888 | path->slots[0]--; | |
1889 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); | |
1890 | if (key.objectid == root->root_key.objectid && | |
1891 | key.type == BTRFS_ROOT_REF_KEY) | |
1892 | ret = -ENOTEMPTY; | |
1893 | } | |
1894 | out: | |
1895 | btrfs_free_path(path); | |
1896 | return ret; | |
1897 | } | |
1898 | ||
ac8e9819 CM |
1899 | static noinline int key_in_sk(struct btrfs_key *key, |
1900 | struct btrfs_ioctl_search_key *sk) | |
1901 | { | |
abc6e134 CM |
1902 | struct btrfs_key test; |
1903 | int ret; | |
1904 | ||
1905 | test.objectid = sk->min_objectid; | |
1906 | test.type = sk->min_type; | |
1907 | test.offset = sk->min_offset; | |
1908 | ||
1909 | ret = btrfs_comp_cpu_keys(key, &test); | |
1910 | if (ret < 0) | |
ac8e9819 | 1911 | return 0; |
abc6e134 CM |
1912 | |
1913 | test.objectid = sk->max_objectid; | |
1914 | test.type = sk->max_type; | |
1915 | test.offset = sk->max_offset; | |
1916 | ||
1917 | ret = btrfs_comp_cpu_keys(key, &test); | |
1918 | if (ret > 0) | |
ac8e9819 CM |
1919 | return 0; |
1920 | return 1; | |
1921 | } | |
1922 | ||
1923 | static noinline int copy_to_sk(struct btrfs_root *root, | |
1924 | struct btrfs_path *path, | |
1925 | struct btrfs_key *key, | |
1926 | struct btrfs_ioctl_search_key *sk, | |
9b6e817d | 1927 | size_t *buf_size, |
ba346b35 | 1928 | char __user *ubuf, |
ac8e9819 CM |
1929 | unsigned long *sk_offset, |
1930 | int *num_found) | |
1931 | { | |
1932 | u64 found_transid; | |
1933 | struct extent_buffer *leaf; | |
1934 | struct btrfs_ioctl_search_header sh; | |
1935 | unsigned long item_off; | |
1936 | unsigned long item_len; | |
1937 | int nritems; | |
1938 | int i; | |
1939 | int slot; | |
ac8e9819 CM |
1940 | int ret = 0; |
1941 | ||
1942 | leaf = path->nodes[0]; | |
1943 | slot = path->slots[0]; | |
1944 | nritems = btrfs_header_nritems(leaf); | |
1945 | ||
1946 | if (btrfs_header_generation(leaf) > sk->max_transid) { | |
1947 | i = nritems; | |
1948 | goto advance_key; | |
1949 | } | |
1950 | found_transid = btrfs_header_generation(leaf); | |
1951 | ||
1952 | for (i = slot; i < nritems; i++) { | |
1953 | item_off = btrfs_item_ptr_offset(leaf, i); | |
1954 | item_len = btrfs_item_size_nr(leaf, i); | |
1955 | ||
03b71c6c GP |
1956 | btrfs_item_key_to_cpu(leaf, key, i); |
1957 | if (!key_in_sk(key, sk)) | |
1958 | continue; | |
1959 | ||
9b6e817d | 1960 | if (sizeof(sh) + item_len > *buf_size) { |
8f5f6178 GH |
1961 | if (*num_found) { |
1962 | ret = 1; | |
1963 | goto out; | |
1964 | } | |
1965 | ||
1966 | /* | |
1967 | * return one empty item back for v1, which does not | |
1968 | * handle -EOVERFLOW | |
1969 | */ | |
1970 | ||
9b6e817d | 1971 | *buf_size = sizeof(sh) + item_len; |
ac8e9819 | 1972 | item_len = 0; |
8f5f6178 GH |
1973 | ret = -EOVERFLOW; |
1974 | } | |
ac8e9819 | 1975 | |
9b6e817d | 1976 | if (sizeof(sh) + item_len + *sk_offset > *buf_size) { |
ac8e9819 | 1977 | ret = 1; |
25c9bc2e | 1978 | goto out; |
ac8e9819 CM |
1979 | } |
1980 | ||
ac8e9819 CM |
1981 | sh.objectid = key->objectid; |
1982 | sh.offset = key->offset; | |
1983 | sh.type = key->type; | |
1984 | sh.len = item_len; | |
1985 | sh.transid = found_transid; | |
1986 | ||
1987 | /* copy search result header */ | |
ba346b35 GH |
1988 | if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { |
1989 | ret = -EFAULT; | |
1990 | goto out; | |
1991 | } | |
1992 | ||
ac8e9819 CM |
1993 | *sk_offset += sizeof(sh); |
1994 | ||
1995 | if (item_len) { | |
ba346b35 | 1996 | char __user *up = ubuf + *sk_offset; |
ac8e9819 | 1997 | /* copy the item */ |
ba346b35 GH |
1998 | if (read_extent_buffer_to_user(leaf, up, |
1999 | item_off, item_len)) { | |
2000 | ret = -EFAULT; | |
2001 | goto out; | |
2002 | } | |
2003 | ||
ac8e9819 | 2004 | *sk_offset += item_len; |
ac8e9819 | 2005 | } |
e2156867 | 2006 | (*num_found)++; |
ac8e9819 | 2007 | |
8f5f6178 GH |
2008 | if (ret) /* -EOVERFLOW from above */ |
2009 | goto out; | |
2010 | ||
25c9bc2e GH |
2011 | if (*num_found >= sk->nr_items) { |
2012 | ret = 1; | |
2013 | goto out; | |
2014 | } | |
ac8e9819 CM |
2015 | } |
2016 | advance_key: | |
abc6e134 CM |
2017 | ret = 0; |
2018 | if (key->offset < (u64)-1 && key->offset < sk->max_offset) | |
ac8e9819 | 2019 | key->offset++; |
abc6e134 CM |
2020 | else if (key->type < (u8)-1 && key->type < sk->max_type) { |
2021 | key->offset = 0; | |
ac8e9819 | 2022 | key->type++; |
abc6e134 CM |
2023 | } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) { |
2024 | key->offset = 0; | |
2025 | key->type = 0; | |
ac8e9819 | 2026 | key->objectid++; |
abc6e134 CM |
2027 | } else |
2028 | ret = 1; | |
25c9bc2e | 2029 | out: |
ba346b35 GH |
2030 | /* |
2031 | * 0: all items from this leaf copied, continue with next | |
2032 | * 1: * more items can be copied, but unused buffer is too small | |
2033 | * * all items were found | |
2034 | * Either way, it will stops the loop which iterates to the next | |
2035 | * leaf | |
2036 | * -EOVERFLOW: item was to large for buffer | |
2037 | * -EFAULT: could not copy extent buffer back to userspace | |
2038 | */ | |
ac8e9819 CM |
2039 | return ret; |
2040 | } | |
2041 | ||
2042 | static noinline int search_ioctl(struct inode *inode, | |
12544442 | 2043 | struct btrfs_ioctl_search_key *sk, |
9b6e817d | 2044 | size_t *buf_size, |
ba346b35 | 2045 | char __user *ubuf) |
ac8e9819 CM |
2046 | { |
2047 | struct btrfs_root *root; | |
2048 | struct btrfs_key key; | |
ac8e9819 | 2049 | struct btrfs_path *path; |
ac8e9819 CM |
2050 | struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info; |
2051 | int ret; | |
2052 | int num_found = 0; | |
2053 | unsigned long sk_offset = 0; | |
2054 | ||
9b6e817d GH |
2055 | if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { |
2056 | *buf_size = sizeof(struct btrfs_ioctl_search_header); | |
12544442 | 2057 | return -EOVERFLOW; |
9b6e817d | 2058 | } |
12544442 | 2059 | |
ac8e9819 CM |
2060 | path = btrfs_alloc_path(); |
2061 | if (!path) | |
2062 | return -ENOMEM; | |
2063 | ||
2064 | if (sk->tree_id == 0) { | |
2065 | /* search the root of the inode that was passed */ | |
2066 | root = BTRFS_I(inode)->root; | |
2067 | } else { | |
2068 | key.objectid = sk->tree_id; | |
2069 | key.type = BTRFS_ROOT_ITEM_KEY; | |
2070 | key.offset = (u64)-1; | |
2071 | root = btrfs_read_fs_root_no_name(info, &key); | |
2072 | if (IS_ERR(root)) { | |
efe120a0 | 2073 | printk(KERN_ERR "BTRFS: could not find root %llu\n", |
ac8e9819 CM |
2074 | sk->tree_id); |
2075 | btrfs_free_path(path); | |
2076 | return -ENOENT; | |
2077 | } | |
2078 | } | |
2079 | ||
2080 | key.objectid = sk->min_objectid; | |
2081 | key.type = sk->min_type; | |
2082 | key.offset = sk->min_offset; | |
2083 | ||
67871254 | 2084 | while (1) { |
6174d3cb | 2085 | ret = btrfs_search_forward(root, &key, path, sk->min_transid); |
ac8e9819 CM |
2086 | if (ret != 0) { |
2087 | if (ret > 0) | |
2088 | ret = 0; | |
2089 | goto err; | |
2090 | } | |
ba346b35 | 2091 | ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf, |
ac8e9819 | 2092 | &sk_offset, &num_found); |
b3b4aa74 | 2093 | btrfs_release_path(path); |
25c9bc2e | 2094 | if (ret) |
ac8e9819 CM |
2095 | break; |
2096 | ||
2097 | } | |
8f5f6178 GH |
2098 | if (ret > 0) |
2099 | ret = 0; | |
ac8e9819 CM |
2100 | err: |
2101 | sk->nr_items = num_found; | |
2102 | btrfs_free_path(path); | |
2103 | return ret; | |
2104 | } | |
2105 | ||
2106 | static noinline int btrfs_ioctl_tree_search(struct file *file, | |
2107 | void __user *argp) | |
2108 | { | |
ba346b35 GH |
2109 | struct btrfs_ioctl_search_args __user *uargs; |
2110 | struct btrfs_ioctl_search_key sk; | |
9b6e817d GH |
2111 | struct inode *inode; |
2112 | int ret; | |
2113 | size_t buf_size; | |
ac8e9819 CM |
2114 | |
2115 | if (!capable(CAP_SYS_ADMIN)) | |
2116 | return -EPERM; | |
2117 | ||
ba346b35 GH |
2118 | uargs = (struct btrfs_ioctl_search_args __user *)argp; |
2119 | ||
2120 | if (copy_from_user(&sk, &uargs->key, sizeof(sk))) | |
2121 | return -EFAULT; | |
ac8e9819 | 2122 | |
ba346b35 | 2123 | buf_size = sizeof(uargs->buf); |
ac8e9819 | 2124 | |
496ad9aa | 2125 | inode = file_inode(file); |
ba346b35 | 2126 | ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); |
8f5f6178 GH |
2127 | |
2128 | /* | |
2129 | * In the origin implementation an overflow is handled by returning a | |
2130 | * search header with a len of zero, so reset ret. | |
2131 | */ | |
2132 | if (ret == -EOVERFLOW) | |
2133 | ret = 0; | |
2134 | ||
ba346b35 | 2135 | if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) |
ac8e9819 | 2136 | ret = -EFAULT; |
ac8e9819 CM |
2137 | return ret; |
2138 | } | |
2139 | ||
cc68a8a5 GH |
2140 | static noinline int btrfs_ioctl_tree_search_v2(struct file *file, |
2141 | void __user *argp) | |
2142 | { | |
2143 | struct btrfs_ioctl_search_args_v2 __user *uarg; | |
2144 | struct btrfs_ioctl_search_args_v2 args; | |
2145 | struct inode *inode; | |
2146 | int ret; | |
2147 | size_t buf_size; | |
2148 | const size_t buf_limit = 16 * 1024 * 1024; | |
2149 | ||
2150 | if (!capable(CAP_SYS_ADMIN)) | |
2151 | return -EPERM; | |
2152 | ||
2153 | /* copy search header and buffer size */ | |
2154 | uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; | |
2155 | if (copy_from_user(&args, uarg, sizeof(args))) | |
2156 | return -EFAULT; | |
2157 | ||
2158 | buf_size = args.buf_size; | |
2159 | ||
2160 | if (buf_size < sizeof(struct btrfs_ioctl_search_header)) | |
2161 | return -EOVERFLOW; | |
2162 | ||
2163 | /* limit result size to 16MB */ | |
2164 | if (buf_size > buf_limit) | |
2165 | buf_size = buf_limit; | |
2166 | ||
2167 | inode = file_inode(file); | |
2168 | ret = search_ioctl(inode, &args.key, &buf_size, | |
2169 | (char *)(&uarg->buf[0])); | |
2170 | if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) | |
2171 | ret = -EFAULT; | |
2172 | else if (ret == -EOVERFLOW && | |
2173 | copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) | |
2174 | ret = -EFAULT; | |
2175 | ||
ac8e9819 CM |
2176 | return ret; |
2177 | } | |
2178 | ||
98d377a0 | 2179 | /* |
ac8e9819 CM |
2180 | * Search INODE_REFs to identify path name of 'dirid' directory |
2181 | * in a 'tree_id' tree. and sets path name to 'name'. | |
2182 | */ | |
98d377a0 TH |
2183 | static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, |
2184 | u64 tree_id, u64 dirid, char *name) | |
2185 | { | |
2186 | struct btrfs_root *root; | |
2187 | struct btrfs_key key; | |
ac8e9819 | 2188 | char *ptr; |
98d377a0 TH |
2189 | int ret = -1; |
2190 | int slot; | |
2191 | int len; | |
2192 | int total_len = 0; | |
2193 | struct btrfs_inode_ref *iref; | |
2194 | struct extent_buffer *l; | |
2195 | struct btrfs_path *path; | |
2196 | ||
2197 | if (dirid == BTRFS_FIRST_FREE_OBJECTID) { | |
2198 | name[0]='\0'; | |
2199 | return 0; | |
2200 | } | |
2201 | ||
2202 | path = btrfs_alloc_path(); | |
2203 | if (!path) | |
2204 | return -ENOMEM; | |
2205 | ||
ac8e9819 | 2206 | ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; |
98d377a0 TH |
2207 | |
2208 | key.objectid = tree_id; | |
2209 | key.type = BTRFS_ROOT_ITEM_KEY; | |
2210 | key.offset = (u64)-1; | |
2211 | root = btrfs_read_fs_root_no_name(info, &key); | |
2212 | if (IS_ERR(root)) { | |
efe120a0 | 2213 | printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id); |
8ad6fcab CM |
2214 | ret = -ENOENT; |
2215 | goto out; | |
98d377a0 TH |
2216 | } |
2217 | ||
2218 | key.objectid = dirid; | |
2219 | key.type = BTRFS_INODE_REF_KEY; | |
8ad6fcab | 2220 | key.offset = (u64)-1; |
98d377a0 | 2221 | |
67871254 | 2222 | while (1) { |
98d377a0 TH |
2223 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
2224 | if (ret < 0) | |
2225 | goto out; | |
18674c6c FDBM |
2226 | else if (ret > 0) { |
2227 | ret = btrfs_previous_item(root, path, dirid, | |
2228 | BTRFS_INODE_REF_KEY); | |
2229 | if (ret < 0) | |
2230 | goto out; | |
2231 | else if (ret > 0) { | |
2232 | ret = -ENOENT; | |
2233 | goto out; | |
2234 | } | |
2235 | } | |
98d377a0 TH |
2236 | |
2237 | l = path->nodes[0]; | |
2238 | slot = path->slots[0]; | |
2239 | btrfs_item_key_to_cpu(l, &key, slot); | |
2240 | ||
98d377a0 TH |
2241 | iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); |
2242 | len = btrfs_inode_ref_name_len(l, iref); | |
2243 | ptr -= len + 1; | |
2244 | total_len += len + 1; | |
a696cf35 FDBM |
2245 | if (ptr < name) { |
2246 | ret = -ENAMETOOLONG; | |
98d377a0 | 2247 | goto out; |
a696cf35 | 2248 | } |
98d377a0 TH |
2249 | |
2250 | *(ptr + len) = '/'; | |
67871254 | 2251 | read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); |
98d377a0 TH |
2252 | |
2253 | if (key.offset == BTRFS_FIRST_FREE_OBJECTID) | |
2254 | break; | |
2255 | ||
b3b4aa74 | 2256 | btrfs_release_path(path); |
98d377a0 | 2257 | key.objectid = key.offset; |
8ad6fcab | 2258 | key.offset = (u64)-1; |
98d377a0 | 2259 | dirid = key.objectid; |
98d377a0 | 2260 | } |
77906a50 | 2261 | memmove(name, ptr, total_len); |
67871254 | 2262 | name[total_len] = '\0'; |
98d377a0 TH |
2263 | ret = 0; |
2264 | out: | |
2265 | btrfs_free_path(path); | |
ac8e9819 CM |
2266 | return ret; |
2267 | } | |
2268 | ||
2269 | static noinline int btrfs_ioctl_ino_lookup(struct file *file, | |
2270 | void __user *argp) | |
2271 | { | |
2272 | struct btrfs_ioctl_ino_lookup_args *args; | |
2273 | struct inode *inode; | |
2274 | int ret; | |
2275 | ||
2276 | if (!capable(CAP_SYS_ADMIN)) | |
2277 | return -EPERM; | |
2278 | ||
2354d08f JL |
2279 | args = memdup_user(argp, sizeof(*args)); |
2280 | if (IS_ERR(args)) | |
2281 | return PTR_ERR(args); | |
c2b96929 | 2282 | |
496ad9aa | 2283 | inode = file_inode(file); |
ac8e9819 | 2284 | |
1b53ac4d CM |
2285 | if (args->treeid == 0) |
2286 | args->treeid = BTRFS_I(inode)->root->root_key.objectid; | |
2287 | ||
ac8e9819 CM |
2288 | ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, |
2289 | args->treeid, args->objectid, | |
2290 | args->name); | |
2291 | ||
2292 | if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) | |
2293 | ret = -EFAULT; | |
2294 | ||
2295 | kfree(args); | |
98d377a0 TH |
2296 | return ret; |
2297 | } | |
2298 | ||
76dda93c YZ |
2299 | static noinline int btrfs_ioctl_snap_destroy(struct file *file, |
2300 | void __user *arg) | |
2301 | { | |
54563d41 | 2302 | struct dentry *parent = file->f_path.dentry; |
76dda93c | 2303 | struct dentry *dentry; |
2b0143b5 | 2304 | struct inode *dir = d_inode(parent); |
76dda93c YZ |
2305 | struct inode *inode; |
2306 | struct btrfs_root *root = BTRFS_I(dir)->root; | |
2307 | struct btrfs_root *dest = NULL; | |
2308 | struct btrfs_ioctl_vol_args *vol_args; | |
2309 | struct btrfs_trans_handle *trans; | |
c58aaad2 | 2310 | struct btrfs_block_rsv block_rsv; |
521e0546 | 2311 | u64 root_flags; |
c58aaad2 | 2312 | u64 qgroup_reserved; |
76dda93c YZ |
2313 | int namelen; |
2314 | int ret; | |
2315 | int err = 0; | |
2316 | ||
76dda93c YZ |
2317 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
2318 | if (IS_ERR(vol_args)) | |
2319 | return PTR_ERR(vol_args); | |
2320 | ||
2321 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
2322 | namelen = strlen(vol_args->name); | |
2323 | if (strchr(vol_args->name, '/') || | |
2324 | strncmp(vol_args->name, "..", namelen) == 0) { | |
2325 | err = -EINVAL; | |
2326 | goto out; | |
2327 | } | |
2328 | ||
a561be71 | 2329 | err = mnt_want_write_file(file); |
76dda93c YZ |
2330 | if (err) |
2331 | goto out; | |
2332 | ||
521e0546 | 2333 | |
5c50c9b8 DS |
2334 | err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT); |
2335 | if (err == -EINTR) | |
e43f998e | 2336 | goto out_drop_write; |
76dda93c YZ |
2337 | dentry = lookup_one_len(vol_args->name, parent, namelen); |
2338 | if (IS_ERR(dentry)) { | |
2339 | err = PTR_ERR(dentry); | |
2340 | goto out_unlock_dir; | |
2341 | } | |
2342 | ||
2b0143b5 | 2343 | if (d_really_is_negative(dentry)) { |
76dda93c YZ |
2344 | err = -ENOENT; |
2345 | goto out_dput; | |
2346 | } | |
2347 | ||
2b0143b5 | 2348 | inode = d_inode(dentry); |
4260f7c7 | 2349 | dest = BTRFS_I(inode)->root; |
67871254 | 2350 | if (!capable(CAP_SYS_ADMIN)) { |
4260f7c7 SW |
2351 | /* |
2352 | * Regular user. Only allow this with a special mount | |
2353 | * option, when the user has write+exec access to the | |
2354 | * subvol root, and when rmdir(2) would have been | |
2355 | * allowed. | |
2356 | * | |
2357 | * Note that this is _not_ check that the subvol is | |
2358 | * empty or doesn't contain data that we wouldn't | |
2359 | * otherwise be able to delete. | |
2360 | * | |
2361 | * Users who want to delete empty subvols should try | |
2362 | * rmdir(2). | |
2363 | */ | |
2364 | err = -EPERM; | |
2365 | if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED)) | |
2366 | goto out_dput; | |
2367 | ||
2368 | /* | |
2369 | * Do not allow deletion if the parent dir is the same | |
2370 | * as the dir to be deleted. That means the ioctl | |
2371 | * must be called on the dentry referencing the root | |
2372 | * of the subvol, not a random directory contained | |
2373 | * within it. | |
2374 | */ | |
2375 | err = -EINVAL; | |
2376 | if (root == dest) | |
2377 | goto out_dput; | |
2378 | ||
2379 | err = inode_permission(inode, MAY_WRITE | MAY_EXEC); | |
2380 | if (err) | |
2381 | goto out_dput; | |
4260f7c7 SW |
2382 | } |
2383 | ||
5c39da5b MX |
2384 | /* check if subvolume may be deleted by a user */ |
2385 | err = btrfs_may_delete(dir, dentry, 1); | |
2386 | if (err) | |
2387 | goto out_dput; | |
2388 | ||
33345d01 | 2389 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { |
76dda93c YZ |
2390 | err = -EINVAL; |
2391 | goto out_dput; | |
2392 | } | |
2393 | ||
76dda93c | 2394 | mutex_lock(&inode->i_mutex); |
521e0546 DS |
2395 | |
2396 | /* | |
2397 | * Don't allow to delete a subvolume with send in progress. This is | |
2398 | * inside the i_mutex so the error handling that has to drop the bit | |
2399 | * again is not run concurrently. | |
2400 | */ | |
2401 | spin_lock(&dest->root_item_lock); | |
c55bfa67 FM |
2402 | root_flags = btrfs_root_flags(&dest->root_item); |
2403 | if (dest->send_in_progress == 0) { | |
2404 | btrfs_set_root_flags(&dest->root_item, | |
521e0546 DS |
2405 | root_flags | BTRFS_ROOT_SUBVOL_DEAD); |
2406 | spin_unlock(&dest->root_item_lock); | |
2407 | } else { | |
2408 | spin_unlock(&dest->root_item_lock); | |
2409 | btrfs_warn(root->fs_info, | |
2410 | "Attempt to delete subvolume %llu during send", | |
c55bfa67 | 2411 | dest->root_key.objectid); |
521e0546 | 2412 | err = -EPERM; |
909e26dc | 2413 | goto out_unlock_inode; |
521e0546 DS |
2414 | } |
2415 | ||
76dda93c YZ |
2416 | down_write(&root->fs_info->subvol_sem); |
2417 | ||
2418 | err = may_destroy_subvol(dest); | |
2419 | if (err) | |
2420 | goto out_up_write; | |
2421 | ||
c58aaad2 MX |
2422 | btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); |
2423 | /* | |
2424 | * One for dir inode, two for dir entries, two for root | |
2425 | * ref/backref. | |
2426 | */ | |
2427 | err = btrfs_subvolume_reserve_metadata(root, &block_rsv, | |
ee3441b4 | 2428 | 5, &qgroup_reserved, true); |
c58aaad2 MX |
2429 | if (err) |
2430 | goto out_up_write; | |
2431 | ||
a22285a6 YZ |
2432 | trans = btrfs_start_transaction(root, 0); |
2433 | if (IS_ERR(trans)) { | |
2434 | err = PTR_ERR(trans); | |
c58aaad2 | 2435 | goto out_release; |
a22285a6 | 2436 | } |
c58aaad2 MX |
2437 | trans->block_rsv = &block_rsv; |
2438 | trans->bytes_reserved = block_rsv.size; | |
a22285a6 | 2439 | |
76dda93c YZ |
2440 | ret = btrfs_unlink_subvol(trans, root, dir, |
2441 | dest->root_key.objectid, | |
2442 | dentry->d_name.name, | |
2443 | dentry->d_name.len); | |
79787eaa JM |
2444 | if (ret) { |
2445 | err = ret; | |
2446 | btrfs_abort_transaction(trans, root, ret); | |
2447 | goto out_end_trans; | |
2448 | } | |
76dda93c YZ |
2449 | |
2450 | btrfs_record_root_in_trans(trans, dest); | |
2451 | ||
2452 | memset(&dest->root_item.drop_progress, 0, | |
2453 | sizeof(dest->root_item.drop_progress)); | |
2454 | dest->root_item.drop_level = 0; | |
2455 | btrfs_set_root_refs(&dest->root_item, 0); | |
2456 | ||
27cdeb70 | 2457 | if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) { |
d68fc57b YZ |
2458 | ret = btrfs_insert_orphan_item(trans, |
2459 | root->fs_info->tree_root, | |
2460 | dest->root_key.objectid); | |
79787eaa JM |
2461 | if (ret) { |
2462 | btrfs_abort_transaction(trans, root, ret); | |
2463 | err = ret; | |
2464 | goto out_end_trans; | |
2465 | } | |
d68fc57b | 2466 | } |
dd5f9615 SB |
2467 | |
2468 | ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, | |
2469 | dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL, | |
2470 | dest->root_key.objectid); | |
2471 | if (ret && ret != -ENOENT) { | |
2472 | btrfs_abort_transaction(trans, root, ret); | |
2473 | err = ret; | |
2474 | goto out_end_trans; | |
2475 | } | |
2476 | if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) { | |
2477 | ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, | |
2478 | dest->root_item.received_uuid, | |
2479 | BTRFS_UUID_KEY_RECEIVED_SUBVOL, | |
2480 | dest->root_key.objectid); | |
2481 | if (ret && ret != -ENOENT) { | |
2482 | btrfs_abort_transaction(trans, root, ret); | |
2483 | err = ret; | |
2484 | goto out_end_trans; | |
2485 | } | |
2486 | } | |
2487 | ||
79787eaa | 2488 | out_end_trans: |
c58aaad2 MX |
2489 | trans->block_rsv = NULL; |
2490 | trans->bytes_reserved = 0; | |
531cb13f | 2491 | ret = btrfs_end_transaction(trans, root); |
79787eaa JM |
2492 | if (ret && !err) |
2493 | err = ret; | |
76dda93c | 2494 | inode->i_flags |= S_DEAD; |
c58aaad2 MX |
2495 | out_release: |
2496 | btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); | |
76dda93c YZ |
2497 | out_up_write: |
2498 | up_write(&root->fs_info->subvol_sem); | |
521e0546 DS |
2499 | if (err) { |
2500 | spin_lock(&dest->root_item_lock); | |
c55bfa67 FM |
2501 | root_flags = btrfs_root_flags(&dest->root_item); |
2502 | btrfs_set_root_flags(&dest->root_item, | |
521e0546 DS |
2503 | root_flags & ~BTRFS_ROOT_SUBVOL_DEAD); |
2504 | spin_unlock(&dest->root_item_lock); | |
2505 | } | |
909e26dc | 2506 | out_unlock_inode: |
76dda93c YZ |
2507 | mutex_unlock(&inode->i_mutex); |
2508 | if (!err) { | |
64ad6c48 | 2509 | d_invalidate(dentry); |
76dda93c YZ |
2510 | btrfs_invalidate_inodes(dest); |
2511 | d_delete(dentry); | |
61155aa0 | 2512 | ASSERT(dest->send_in_progress == 0); |
fa6ac876 LB |
2513 | |
2514 | /* the last ref */ | |
57cdc8db DS |
2515 | if (dest->ino_cache_inode) { |
2516 | iput(dest->ino_cache_inode); | |
2517 | dest->ino_cache_inode = NULL; | |
fa6ac876 | 2518 | } |
76dda93c YZ |
2519 | } |
2520 | out_dput: | |
2521 | dput(dentry); | |
2522 | out_unlock_dir: | |
2523 | mutex_unlock(&dir->i_mutex); | |
e43f998e | 2524 | out_drop_write: |
2a79f17e | 2525 | mnt_drop_write_file(file); |
76dda93c YZ |
2526 | out: |
2527 | kfree(vol_args); | |
2528 | return err; | |
2529 | } | |
2530 | ||
1e701a32 | 2531 | static int btrfs_ioctl_defrag(struct file *file, void __user *argp) |
f46b5a66 | 2532 | { |
496ad9aa | 2533 | struct inode *inode = file_inode(file); |
f46b5a66 | 2534 | struct btrfs_root *root = BTRFS_I(inode)->root; |
1e701a32 | 2535 | struct btrfs_ioctl_defrag_range_args *range; |
c146afad YZ |
2536 | int ret; |
2537 | ||
25122d15 ID |
2538 | ret = mnt_want_write_file(file); |
2539 | if (ret) | |
2540 | return ret; | |
b83cc969 | 2541 | |
25122d15 ID |
2542 | if (btrfs_root_readonly(root)) { |
2543 | ret = -EROFS; | |
2544 | goto out; | |
5ac00add | 2545 | } |
f46b5a66 CH |
2546 | |
2547 | switch (inode->i_mode & S_IFMT) { | |
2548 | case S_IFDIR: | |
e441d54d CM |
2549 | if (!capable(CAP_SYS_ADMIN)) { |
2550 | ret = -EPERM; | |
2551 | goto out; | |
2552 | } | |
de78b51a | 2553 | ret = btrfs_defrag_root(root); |
8929ecfa YZ |
2554 | if (ret) |
2555 | goto out; | |
de78b51a | 2556 | ret = btrfs_defrag_root(root->fs_info->extent_root); |
f46b5a66 CH |
2557 | break; |
2558 | case S_IFREG: | |
e441d54d CM |
2559 | if (!(file->f_mode & FMODE_WRITE)) { |
2560 | ret = -EINVAL; | |
2561 | goto out; | |
2562 | } | |
1e701a32 CM |
2563 | |
2564 | range = kzalloc(sizeof(*range), GFP_KERNEL); | |
2565 | if (!range) { | |
2566 | ret = -ENOMEM; | |
2567 | goto out; | |
2568 | } | |
2569 | ||
2570 | if (argp) { | |
2571 | if (copy_from_user(range, argp, | |
2572 | sizeof(*range))) { | |
2573 | ret = -EFAULT; | |
2574 | kfree(range); | |
683be16e | 2575 | goto out; |
1e701a32 CM |
2576 | } |
2577 | /* compression requires us to start the IO */ | |
2578 | if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { | |
2579 | range->flags |= BTRFS_DEFRAG_RANGE_START_IO; | |
2580 | range->extent_thresh = (u32)-1; | |
2581 | } | |
2582 | } else { | |
2583 | /* the rest are all set to zero by kzalloc */ | |
2584 | range->len = (u64)-1; | |
2585 | } | |
496ad9aa | 2586 | ret = btrfs_defrag_file(file_inode(file), file, |
4cb5300b CM |
2587 | range, 0, 0); |
2588 | if (ret > 0) | |
2589 | ret = 0; | |
1e701a32 | 2590 | kfree(range); |
f46b5a66 | 2591 | break; |
8929ecfa YZ |
2592 | default: |
2593 | ret = -EINVAL; | |
f46b5a66 | 2594 | } |
e441d54d | 2595 | out: |
25122d15 | 2596 | mnt_drop_write_file(file); |
e441d54d | 2597 | return ret; |
f46b5a66 CH |
2598 | } |
2599 | ||
b2950863 | 2600 | static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) |
f46b5a66 CH |
2601 | { |
2602 | struct btrfs_ioctl_vol_args *vol_args; | |
2603 | int ret; | |
2604 | ||
e441d54d CM |
2605 | if (!capable(CAP_SYS_ADMIN)) |
2606 | return -EPERM; | |
2607 | ||
5ac00add SB |
2608 | if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, |
2609 | 1)) { | |
e57138b3 | 2610 | return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
c9e9f97b ID |
2611 | } |
2612 | ||
5ac00add | 2613 | mutex_lock(&root->fs_info->volume_mutex); |
dae7b665 | 2614 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
2615 | if (IS_ERR(vol_args)) { |
2616 | ret = PTR_ERR(vol_args); | |
2617 | goto out; | |
2618 | } | |
f46b5a66 | 2619 | |
5516e595 | 2620 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
f46b5a66 CH |
2621 | ret = btrfs_init_new_device(root, vol_args->name); |
2622 | ||
43d20761 AJ |
2623 | if (!ret) |
2624 | btrfs_info(root->fs_info, "disk added %s",vol_args->name); | |
2625 | ||
f46b5a66 | 2626 | kfree(vol_args); |
c9e9f97b ID |
2627 | out: |
2628 | mutex_unlock(&root->fs_info->volume_mutex); | |
5ac00add | 2629 | atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); |
f46b5a66 CH |
2630 | return ret; |
2631 | } | |
2632 | ||
da24927b | 2633 | static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) |
f46b5a66 | 2634 | { |
496ad9aa | 2635 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
f46b5a66 CH |
2636 | struct btrfs_ioctl_vol_args *vol_args; |
2637 | int ret; | |
2638 | ||
e441d54d CM |
2639 | if (!capable(CAP_SYS_ADMIN)) |
2640 | return -EPERM; | |
2641 | ||
da24927b MX |
2642 | ret = mnt_want_write_file(file); |
2643 | if (ret) | |
2644 | return ret; | |
c146afad | 2645 | |
dae7b665 | 2646 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
2647 | if (IS_ERR(vol_args)) { |
2648 | ret = PTR_ERR(vol_args); | |
c47ca32d | 2649 | goto err_drop; |
c9e9f97b | 2650 | } |
f46b5a66 | 2651 | |
5516e595 | 2652 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
f46b5a66 | 2653 | |
183860f6 AJ |
2654 | if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, |
2655 | 1)) { | |
2656 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; | |
2657 | goto out; | |
2658 | } | |
2659 | ||
2660 | mutex_lock(&root->fs_info->volume_mutex); | |
2661 | ret = btrfs_rm_device(root, vol_args->name); | |
c9e9f97b | 2662 | mutex_unlock(&root->fs_info->volume_mutex); |
5ac00add | 2663 | atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); |
183860f6 | 2664 | |
ec95d491 AJ |
2665 | if (!ret) |
2666 | btrfs_info(root->fs_info, "disk deleted %s",vol_args->name); | |
2667 | ||
183860f6 AJ |
2668 | out: |
2669 | kfree(vol_args); | |
c47ca32d | 2670 | err_drop: |
4ac20c70 | 2671 | mnt_drop_write_file(file); |
f46b5a66 CH |
2672 | return ret; |
2673 | } | |
2674 | ||
475f6387 JS |
2675 | static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg) |
2676 | { | |
027ed2f0 | 2677 | struct btrfs_ioctl_fs_info_args *fi_args; |
475f6387 JS |
2678 | struct btrfs_device *device; |
2679 | struct btrfs_device *next; | |
2680 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
027ed2f0 | 2681 | int ret = 0; |
475f6387 | 2682 | |
027ed2f0 LZ |
2683 | fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); |
2684 | if (!fi_args) | |
2685 | return -ENOMEM; | |
2686 | ||
f7171750 | 2687 | mutex_lock(&fs_devices->device_list_mutex); |
027ed2f0 LZ |
2688 | fi_args->num_devices = fs_devices->num_devices; |
2689 | memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid)); | |
475f6387 | 2690 | |
475f6387 | 2691 | list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) { |
027ed2f0 LZ |
2692 | if (device->devid > fi_args->max_id) |
2693 | fi_args->max_id = device->devid; | |
475f6387 JS |
2694 | } |
2695 | mutex_unlock(&fs_devices->device_list_mutex); | |
2696 | ||
80a773fb DS |
2697 | fi_args->nodesize = root->fs_info->super_copy->nodesize; |
2698 | fi_args->sectorsize = root->fs_info->super_copy->sectorsize; | |
2699 | fi_args->clone_alignment = root->fs_info->super_copy->sectorsize; | |
2700 | ||
027ed2f0 LZ |
2701 | if (copy_to_user(arg, fi_args, sizeof(*fi_args))) |
2702 | ret = -EFAULT; | |
475f6387 | 2703 | |
027ed2f0 LZ |
2704 | kfree(fi_args); |
2705 | return ret; | |
475f6387 JS |
2706 | } |
2707 | ||
2708 | static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg) | |
2709 | { | |
2710 | struct btrfs_ioctl_dev_info_args *di_args; | |
2711 | struct btrfs_device *dev; | |
2712 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
2713 | int ret = 0; | |
2714 | char *s_uuid = NULL; | |
475f6387 | 2715 | |
475f6387 JS |
2716 | di_args = memdup_user(arg, sizeof(*di_args)); |
2717 | if (IS_ERR(di_args)) | |
2718 | return PTR_ERR(di_args); | |
2719 | ||
dd5f9615 | 2720 | if (!btrfs_is_empty_uuid(di_args->uuid)) |
475f6387 JS |
2721 | s_uuid = di_args->uuid; |
2722 | ||
2723 | mutex_lock(&fs_devices->device_list_mutex); | |
aa1b8cd4 | 2724 | dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL); |
475f6387 JS |
2725 | |
2726 | if (!dev) { | |
2727 | ret = -ENODEV; | |
2728 | goto out; | |
2729 | } | |
2730 | ||
2731 | di_args->devid = dev->devid; | |
7cc8e58d MX |
2732 | di_args->bytes_used = btrfs_device_get_bytes_used(dev); |
2733 | di_args->total_bytes = btrfs_device_get_total_bytes(dev); | |
475f6387 | 2734 | memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); |
a27202fb | 2735 | if (dev->name) { |
606686ee JB |
2736 | struct rcu_string *name; |
2737 | ||
2738 | rcu_read_lock(); | |
2739 | name = rcu_dereference(dev->name); | |
2740 | strncpy(di_args->path, name->str, sizeof(di_args->path)); | |
2741 | rcu_read_unlock(); | |
a27202fb JM |
2742 | di_args->path[sizeof(di_args->path) - 1] = 0; |
2743 | } else { | |
99ba55ad | 2744 | di_args->path[0] = '\0'; |
a27202fb | 2745 | } |
475f6387 JS |
2746 | |
2747 | out: | |
55793c0d | 2748 | mutex_unlock(&fs_devices->device_list_mutex); |
475f6387 JS |
2749 | if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) |
2750 | ret = -EFAULT; | |
2751 | ||
2752 | kfree(di_args); | |
2753 | return ret; | |
2754 | } | |
2755 | ||
416161db MF |
2756 | static struct page *extent_same_get_page(struct inode *inode, u64 off) |
2757 | { | |
2758 | struct page *page; | |
2759 | pgoff_t index; | |
2760 | struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree; | |
2761 | ||
2762 | index = off >> PAGE_CACHE_SHIFT; | |
2763 | ||
2764 | page = grab_cache_page(inode->i_mapping, index); | |
2765 | if (!page) | |
2766 | return NULL; | |
2767 | ||
2768 | if (!PageUptodate(page)) { | |
2769 | if (extent_read_full_page_nolock(tree, page, btrfs_get_extent, | |
2770 | 0)) | |
2771 | return NULL; | |
2772 | lock_page(page); | |
2773 | if (!PageUptodate(page)) { | |
2774 | unlock_page(page); | |
2775 | page_cache_release(page); | |
2776 | return NULL; | |
2777 | } | |
2778 | } | |
2779 | unlock_page(page); | |
2780 | ||
2781 | return page; | |
2782 | } | |
2783 | ||
77fe20dc MF |
2784 | static inline void lock_extent_range(struct inode *inode, u64 off, u64 len) |
2785 | { | |
2786 | /* do any pending delalloc/csum calc on src, one way or | |
2787 | another, and lock file content */ | |
2788 | while (1) { | |
2789 | struct btrfs_ordered_extent *ordered; | |
2790 | lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); | |
2791 | ordered = btrfs_lookup_first_ordered_extent(inode, | |
2792 | off + len - 1); | |
ff5df9b8 FM |
2793 | if ((!ordered || |
2794 | ordered->file_offset + ordered->len <= off || | |
2795 | ordered->file_offset >= off + len) && | |
77fe20dc | 2796 | !test_range_bit(&BTRFS_I(inode)->io_tree, off, |
ff5df9b8 FM |
2797 | off + len - 1, EXTENT_DELALLOC, 0, NULL)) { |
2798 | if (ordered) | |
2799 | btrfs_put_ordered_extent(ordered); | |
77fe20dc | 2800 | break; |
ff5df9b8 | 2801 | } |
77fe20dc MF |
2802 | unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); |
2803 | if (ordered) | |
2804 | btrfs_put_ordered_extent(ordered); | |
2805 | btrfs_wait_ordered_range(inode, off, len); | |
2806 | } | |
2807 | } | |
2808 | ||
416161db MF |
2809 | static void btrfs_double_unlock(struct inode *inode1, u64 loff1, |
2810 | struct inode *inode2, u64 loff2, u64 len) | |
2811 | { | |
2812 | unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); | |
2813 | unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); | |
2814 | ||
2815 | mutex_unlock(&inode1->i_mutex); | |
2816 | mutex_unlock(&inode2->i_mutex); | |
2817 | } | |
2818 | ||
2819 | static void btrfs_double_lock(struct inode *inode1, u64 loff1, | |
2820 | struct inode *inode2, u64 loff2, u64 len) | |
2821 | { | |
2822 | if (inode1 < inode2) { | |
2823 | swap(inode1, inode2); | |
2824 | swap(loff1, loff2); | |
2825 | } | |
2826 | ||
2827 | mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT); | |
2828 | lock_extent_range(inode1, loff1, len); | |
2829 | if (inode1 != inode2) { | |
2830 | mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD); | |
2831 | lock_extent_range(inode2, loff2, len); | |
2832 | } | |
2833 | } | |
2834 | ||
2835 | static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst, | |
2836 | u64 dst_loff, u64 len) | |
2837 | { | |
2838 | int ret = 0; | |
2839 | struct page *src_page, *dst_page; | |
2840 | unsigned int cmp_len = PAGE_CACHE_SIZE; | |
2841 | void *addr, *dst_addr; | |
2842 | ||
2843 | while (len) { | |
2844 | if (len < PAGE_CACHE_SIZE) | |
2845 | cmp_len = len; | |
2846 | ||
2847 | src_page = extent_same_get_page(src, loff); | |
2848 | if (!src_page) | |
2849 | return -EINVAL; | |
2850 | dst_page = extent_same_get_page(dst, dst_loff); | |
2851 | if (!dst_page) { | |
2852 | page_cache_release(src_page); | |
2853 | return -EINVAL; | |
2854 | } | |
2855 | addr = kmap_atomic(src_page); | |
2856 | dst_addr = kmap_atomic(dst_page); | |
2857 | ||
2858 | flush_dcache_page(src_page); | |
2859 | flush_dcache_page(dst_page); | |
2860 | ||
2861 | if (memcmp(addr, dst_addr, cmp_len)) | |
2862 | ret = BTRFS_SAME_DATA_DIFFERS; | |
2863 | ||
2864 | kunmap_atomic(addr); | |
2865 | kunmap_atomic(dst_addr); | |
2866 | page_cache_release(src_page); | |
2867 | page_cache_release(dst_page); | |
2868 | ||
2869 | if (ret) | |
2870 | break; | |
2871 | ||
2872 | loff += cmp_len; | |
2873 | dst_loff += cmp_len; | |
2874 | len -= cmp_len; | |
2875 | } | |
2876 | ||
2877 | return ret; | |
2878 | } | |
2879 | ||
2880 | static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len) | |
2881 | { | |
2882 | u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; | |
2883 | ||
2884 | if (off + len > inode->i_size || off + len < off) | |
2885 | return -EINVAL; | |
2886 | /* Check that we are block aligned - btrfs_clone() requires this */ | |
2887 | if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) | |
2888 | return -EINVAL; | |
2889 | ||
2890 | return 0; | |
2891 | } | |
2892 | ||
2893 | static int btrfs_extent_same(struct inode *src, u64 loff, u64 len, | |
2894 | struct inode *dst, u64 dst_loff) | |
2895 | { | |
2896 | int ret; | |
2897 | ||
2898 | /* | |
2899 | * btrfs_clone() can't handle extents in the same file | |
2900 | * yet. Once that works, we can drop this check and replace it | |
2901 | * with a check for the same inode, but overlapping extents. | |
2902 | */ | |
2903 | if (src == dst) | |
2904 | return -EINVAL; | |
2905 | ||
113e8283 FM |
2906 | if (len == 0) |
2907 | return 0; | |
2908 | ||
416161db MF |
2909 | btrfs_double_lock(src, loff, dst, dst_loff, len); |
2910 | ||
2911 | ret = extent_same_check_offsets(src, loff, len); | |
2912 | if (ret) | |
2913 | goto out_unlock; | |
2914 | ||
2915 | ret = extent_same_check_offsets(dst, dst_loff, len); | |
2916 | if (ret) | |
2917 | goto out_unlock; | |
2918 | ||
2919 | /* don't make the dst file partly checksummed */ | |
2920 | if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != | |
2921 | (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { | |
2922 | ret = -EINVAL; | |
2923 | goto out_unlock; | |
2924 | } | |
2925 | ||
2926 | ret = btrfs_cmp_data(src, loff, dst, dst_loff, len); | |
2927 | if (ret == 0) | |
2928 | ret = btrfs_clone(src, dst, loff, len, len, dst_loff); | |
2929 | ||
2930 | out_unlock: | |
2931 | btrfs_double_unlock(src, loff, dst, dst_loff, len); | |
2932 | ||
2933 | return ret; | |
2934 | } | |
2935 | ||
2936 | #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024) | |
2937 | ||
2938 | static long btrfs_ioctl_file_extent_same(struct file *file, | |
1c1c8747 | 2939 | struct btrfs_ioctl_same_args __user *argp) |
416161db | 2940 | { |
cbf8b8ca MF |
2941 | struct btrfs_ioctl_same_args *same; |
2942 | struct btrfs_ioctl_same_extent_info *info; | |
1c1c8747 | 2943 | struct inode *src = file_inode(file); |
416161db MF |
2944 | u64 off; |
2945 | u64 len; | |
2946 | int i; | |
2947 | int ret; | |
cbf8b8ca | 2948 | unsigned long size; |
416161db MF |
2949 | u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; |
2950 | bool is_admin = capable(CAP_SYS_ADMIN); | |
1c1c8747 | 2951 | u16 count; |
416161db MF |
2952 | |
2953 | if (!(file->f_mode & FMODE_READ)) | |
2954 | return -EINVAL; | |
2955 | ||
2956 | ret = mnt_want_write_file(file); | |
2957 | if (ret) | |
2958 | return ret; | |
2959 | ||
1c1c8747 | 2960 | if (get_user(count, &argp->dest_count)) { |
416161db MF |
2961 | ret = -EFAULT; |
2962 | goto out; | |
2963 | } | |
2964 | ||
1c1c8747 | 2965 | size = offsetof(struct btrfs_ioctl_same_args __user, info[count]); |
cbf8b8ca | 2966 | |
1c1c8747 | 2967 | same = memdup_user(argp, size); |
cbf8b8ca | 2968 | |
229eed43 GB |
2969 | if (IS_ERR(same)) { |
2970 | ret = PTR_ERR(same); | |
cbf8b8ca MF |
2971 | goto out; |
2972 | } | |
2973 | ||
2974 | off = same->logical_offset; | |
2975 | len = same->length; | |
416161db MF |
2976 | |
2977 | /* | |
2978 | * Limit the total length we will dedupe for each operation. | |
2979 | * This is intended to bound the total time spent in this | |
2980 | * ioctl to something sane. | |
2981 | */ | |
2982 | if (len > BTRFS_MAX_DEDUPE_LEN) | |
2983 | len = BTRFS_MAX_DEDUPE_LEN; | |
2984 | ||
2985 | if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) { | |
2986 | /* | |
2987 | * Btrfs does not support blocksize < page_size. As a | |
2988 | * result, btrfs_cmp_data() won't correctly handle | |
2989 | * this situation without an update. | |
2990 | */ | |
2991 | ret = -EINVAL; | |
2992 | goto out; | |
2993 | } | |
2994 | ||
2995 | ret = -EISDIR; | |
2996 | if (S_ISDIR(src->i_mode)) | |
2997 | goto out; | |
2998 | ||
2999 | ret = -EACCES; | |
3000 | if (!S_ISREG(src->i_mode)) | |
3001 | goto out; | |
3002 | ||
cbf8b8ca | 3003 | /* pre-format output fields to sane values */ |
1c1c8747 | 3004 | for (i = 0; i < count; i++) { |
cbf8b8ca MF |
3005 | same->info[i].bytes_deduped = 0ULL; |
3006 | same->info[i].status = 0; | |
3007 | } | |
416161db | 3008 | |
1c1c8747 AV |
3009 | for (i = 0, info = same->info; i < count; i++, info++) { |
3010 | struct inode *dst; | |
3011 | struct fd dst_file = fdget(info->fd); | |
3012 | if (!dst_file.file) { | |
cbf8b8ca | 3013 | info->status = -EBADF; |
1c1c8747 | 3014 | continue; |
416161db | 3015 | } |
1c1c8747 | 3016 | dst = file_inode(dst_file.file); |
416161db | 3017 | |
1c1c8747 | 3018 | if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) { |
cbf8b8ca | 3019 | info->status = -EINVAL; |
1c1c8747 AV |
3020 | } else if (file->f_path.mnt != dst_file.file->f_path.mnt) { |
3021 | info->status = -EXDEV; | |
3022 | } else if (S_ISDIR(dst->i_mode)) { | |
cbf8b8ca | 3023 | info->status = -EISDIR; |
1c1c8747 | 3024 | } else if (!S_ISREG(dst->i_mode)) { |
cbf8b8ca | 3025 | info->status = -EACCES; |
1c1c8747 AV |
3026 | } else { |
3027 | info->status = btrfs_extent_same(src, off, len, dst, | |
3028 | info->logical_offset); | |
3029 | if (info->status == 0) | |
3030 | info->bytes_deduped += len; | |
416161db | 3031 | } |
1c1c8747 | 3032 | fdput(dst_file); |
416161db MF |
3033 | } |
3034 | ||
cbf8b8ca MF |
3035 | ret = copy_to_user(argp, same, size); |
3036 | if (ret) | |
3037 | ret = -EFAULT; | |
3038 | ||
416161db MF |
3039 | out: |
3040 | mnt_drop_write_file(file); | |
3041 | return ret; | |
3042 | } | |
3043 | ||
fcebe456 JB |
3044 | /* Helper to check and see if this root currently has a ref on the given disk |
3045 | * bytenr. If it does then we need to update the quota for this root. This | |
3046 | * doesn't do anything if quotas aren't enabled. | |
3047 | */ | |
3048 | static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, | |
3049 | u64 disko) | |
3050 | { | |
3284da7b | 3051 | struct seq_list tree_mod_seq_elem = SEQ_LIST_INIT(tree_mod_seq_elem); |
fcebe456 JB |
3052 | struct ulist *roots; |
3053 | struct ulist_iterator uiter; | |
3054 | struct ulist_node *root_node = NULL; | |
3055 | int ret; | |
3056 | ||
3057 | if (!root->fs_info->quota_enabled) | |
3058 | return 1; | |
3059 | ||
3060 | btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem); | |
3061 | ret = btrfs_find_all_roots(trans, root->fs_info, disko, | |
3062 | tree_mod_seq_elem.seq, &roots); | |
3063 | if (ret < 0) | |
3064 | goto out; | |
3065 | ret = 0; | |
3066 | ULIST_ITER_INIT(&uiter); | |
3067 | while ((root_node = ulist_next(roots, &uiter))) { | |
3068 | if (root_node->val == root->objectid) { | |
3069 | ret = 1; | |
3070 | break; | |
3071 | } | |
3072 | } | |
3073 | ulist_free(roots); | |
3074 | out: | |
3075 | btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem); | |
3076 | return ret; | |
3077 | } | |
3078 | ||
f82a9901 FM |
3079 | static int clone_finish_inode_update(struct btrfs_trans_handle *trans, |
3080 | struct inode *inode, | |
3081 | u64 endoff, | |
3082 | const u64 destoff, | |
3083 | const u64 olen) | |
3084 | { | |
3085 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
3086 | int ret; | |
3087 | ||
3088 | inode_inc_iversion(inode); | |
3089 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; | |
3090 | /* | |
3091 | * We round up to the block size at eof when determining which | |
3092 | * extents to clone above, but shouldn't round up the file size. | |
3093 | */ | |
3094 | if (endoff > destoff + olen) | |
3095 | endoff = destoff + olen; | |
3096 | if (endoff > inode->i_size) | |
3097 | btrfs_i_size_write(inode, endoff); | |
3098 | ||
3099 | ret = btrfs_update_inode(trans, root, inode); | |
3100 | if (ret) { | |
3101 | btrfs_abort_transaction(trans, root, ret); | |
3102 | btrfs_end_transaction(trans, root); | |
3103 | goto out; | |
3104 | } | |
3105 | ret = btrfs_end_transaction(trans, root); | |
3106 | out: | |
3107 | return ret; | |
3108 | } | |
3109 | ||
7ffbb598 FM |
3110 | static void clone_update_extent_map(struct inode *inode, |
3111 | const struct btrfs_trans_handle *trans, | |
3112 | const struct btrfs_path *path, | |
7ffbb598 FM |
3113 | const u64 hole_offset, |
3114 | const u64 hole_len) | |
3115 | { | |
3116 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
3117 | struct extent_map *em; | |
3118 | int ret; | |
3119 | ||
3120 | em = alloc_extent_map(); | |
3121 | if (!em) { | |
3122 | set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, | |
3123 | &BTRFS_I(inode)->runtime_flags); | |
3124 | return; | |
3125 | } | |
3126 | ||
14f59796 FM |
3127 | if (path) { |
3128 | struct btrfs_file_extent_item *fi; | |
3129 | ||
3130 | fi = btrfs_item_ptr(path->nodes[0], path->slots[0], | |
3131 | struct btrfs_file_extent_item); | |
7ffbb598 FM |
3132 | btrfs_extent_item_to_extent_map(inode, path, fi, false, em); |
3133 | em->generation = -1; | |
3134 | if (btrfs_file_extent_type(path->nodes[0], fi) == | |
3135 | BTRFS_FILE_EXTENT_INLINE) | |
3136 | set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, | |
3137 | &BTRFS_I(inode)->runtime_flags); | |
3138 | } else { | |
3139 | em->start = hole_offset; | |
3140 | em->len = hole_len; | |
3141 | em->ram_bytes = em->len; | |
3142 | em->orig_start = hole_offset; | |
3143 | em->block_start = EXTENT_MAP_HOLE; | |
3144 | em->block_len = 0; | |
3145 | em->orig_block_len = 0; | |
3146 | em->compress_type = BTRFS_COMPRESS_NONE; | |
3147 | em->generation = trans->transid; | |
3148 | } | |
3149 | ||
3150 | while (1) { | |
3151 | write_lock(&em_tree->lock); | |
3152 | ret = add_extent_mapping(em_tree, em, 1); | |
3153 | write_unlock(&em_tree->lock); | |
3154 | if (ret != -EEXIST) { | |
3155 | free_extent_map(em); | |
3156 | break; | |
3157 | } | |
3158 | btrfs_drop_extent_cache(inode, em->start, | |
3159 | em->start + em->len - 1, 0); | |
3160 | } | |
3161 | ||
ee39b432 | 3162 | if (ret) |
7ffbb598 FM |
3163 | set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, |
3164 | &BTRFS_I(inode)->runtime_flags); | |
3165 | } | |
3166 | ||
32b7c687 MF |
3167 | /** |
3168 | * btrfs_clone() - clone a range from inode file to another | |
3169 | * | |
3170 | * @src: Inode to clone from | |
3171 | * @inode: Inode to clone to | |
3172 | * @off: Offset within source to start clone from | |
3173 | * @olen: Original length, passed by user, of range to clone | |
3174 | * @olen_aligned: Block-aligned value of olen, extent_same uses | |
3175 | * identical values here | |
3176 | * @destoff: Offset within @inode to start clone | |
3177 | */ | |
3178 | static int btrfs_clone(struct inode *src, struct inode *inode, | |
f82a9901 FM |
3179 | const u64 off, const u64 olen, const u64 olen_aligned, |
3180 | const u64 destoff) | |
f46b5a66 | 3181 | { |
f46b5a66 | 3182 | struct btrfs_root *root = BTRFS_I(inode)->root; |
32b7c687 | 3183 | struct btrfs_path *path = NULL; |
f46b5a66 | 3184 | struct extent_buffer *leaf; |
32b7c687 MF |
3185 | struct btrfs_trans_handle *trans; |
3186 | char *buf = NULL; | |
ae01a0ab | 3187 | struct btrfs_key key; |
f46b5a66 CH |
3188 | u32 nritems; |
3189 | int slot; | |
ae01a0ab | 3190 | int ret; |
fcebe456 | 3191 | int no_quota; |
f82a9901 | 3192 | const u64 len = olen_aligned; |
fcebe456 | 3193 | u64 last_disko = 0; |
f82a9901 | 3194 | u64 last_dest_end = destoff; |
ae01a0ab YZ |
3195 | |
3196 | ret = -ENOMEM; | |
707e8a07 | 3197 | buf = vmalloc(root->nodesize); |
ae01a0ab | 3198 | if (!buf) |
32b7c687 | 3199 | return ret; |
ae01a0ab YZ |
3200 | |
3201 | path = btrfs_alloc_path(); | |
3202 | if (!path) { | |
3203 | vfree(buf); | |
32b7c687 | 3204 | return ret; |
d525e8ab LZ |
3205 | } |
3206 | ||
32b7c687 | 3207 | path->reada = 2; |
c5c9cd4d | 3208 | /* clone data */ |
33345d01 | 3209 | key.objectid = btrfs_ino(src); |
ae01a0ab | 3210 | key.type = BTRFS_EXTENT_DATA_KEY; |
2c463823 | 3211 | key.offset = off; |
f46b5a66 CH |
3212 | |
3213 | while (1) { | |
de249e66 | 3214 | u64 next_key_min_offset = key.offset + 1; |
df858e76 | 3215 | |
f46b5a66 CH |
3216 | /* |
3217 | * note the key will change type as we walk through the | |
3218 | * tree. | |
3219 | */ | |
e4355f34 | 3220 | path->leave_spinning = 1; |
362a20c5 DS |
3221 | ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, |
3222 | 0, 0); | |
f46b5a66 CH |
3223 | if (ret < 0) |
3224 | goto out; | |
2c463823 FM |
3225 | /* |
3226 | * First search, if no extent item that starts at offset off was | |
3227 | * found but the previous item is an extent item, it's possible | |
3228 | * it might overlap our target range, therefore process it. | |
3229 | */ | |
3230 | if (key.offset == off && ret > 0 && path->slots[0] > 0) { | |
3231 | btrfs_item_key_to_cpu(path->nodes[0], &key, | |
3232 | path->slots[0] - 1); | |
3233 | if (key.type == BTRFS_EXTENT_DATA_KEY) | |
3234 | path->slots[0]--; | |
3235 | } | |
f46b5a66 | 3236 | |
ae01a0ab | 3237 | nritems = btrfs_header_nritems(path->nodes[0]); |
e4355f34 | 3238 | process_slot: |
fcebe456 | 3239 | no_quota = 1; |
ae01a0ab | 3240 | if (path->slots[0] >= nritems) { |
362a20c5 | 3241 | ret = btrfs_next_leaf(BTRFS_I(src)->root, path); |
f46b5a66 CH |
3242 | if (ret < 0) |
3243 | goto out; | |
3244 | if (ret > 0) | |
3245 | break; | |
ae01a0ab | 3246 | nritems = btrfs_header_nritems(path->nodes[0]); |
f46b5a66 CH |
3247 | } |
3248 | leaf = path->nodes[0]; | |
3249 | slot = path->slots[0]; | |
f46b5a66 | 3250 | |
ae01a0ab | 3251 | btrfs_item_key_to_cpu(leaf, &key, slot); |
962a298f | 3252 | if (key.type > BTRFS_EXTENT_DATA_KEY || |
33345d01 | 3253 | key.objectid != btrfs_ino(src)) |
f46b5a66 CH |
3254 | break; |
3255 | ||
962a298f | 3256 | if (key.type == BTRFS_EXTENT_DATA_KEY) { |
c5c9cd4d SW |
3257 | struct btrfs_file_extent_item *extent; |
3258 | int type; | |
31840ae1 ZY |
3259 | u32 size; |
3260 | struct btrfs_key new_key; | |
c5c9cd4d SW |
3261 | u64 disko = 0, diskl = 0; |
3262 | u64 datao = 0, datal = 0; | |
3263 | u8 comp; | |
f82a9901 | 3264 | u64 drop_start; |
31840ae1 | 3265 | |
c5c9cd4d SW |
3266 | extent = btrfs_item_ptr(leaf, slot, |
3267 | struct btrfs_file_extent_item); | |
3268 | comp = btrfs_file_extent_compression(leaf, extent); | |
3269 | type = btrfs_file_extent_type(leaf, extent); | |
c8a894d7 CM |
3270 | if (type == BTRFS_FILE_EXTENT_REG || |
3271 | type == BTRFS_FILE_EXTENT_PREALLOC) { | |
d397712b CM |
3272 | disko = btrfs_file_extent_disk_bytenr(leaf, |
3273 | extent); | |
3274 | diskl = btrfs_file_extent_disk_num_bytes(leaf, | |
3275 | extent); | |
c5c9cd4d | 3276 | datao = btrfs_file_extent_offset(leaf, extent); |
d397712b CM |
3277 | datal = btrfs_file_extent_num_bytes(leaf, |
3278 | extent); | |
c5c9cd4d SW |
3279 | } else if (type == BTRFS_FILE_EXTENT_INLINE) { |
3280 | /* take upper bound, may be compressed */ | |
3281 | datal = btrfs_file_extent_ram_bytes(leaf, | |
3282 | extent); | |
3283 | } | |
31840ae1 | 3284 | |
2c463823 FM |
3285 | /* |
3286 | * The first search might have left us at an extent | |
3287 | * item that ends before our target range's start, can | |
3288 | * happen if we have holes and NO_HOLES feature enabled. | |
3289 | */ | |
3290 | if (key.offset + datal <= off) { | |
e4355f34 FDBM |
3291 | path->slots[0]++; |
3292 | goto process_slot; | |
2c463823 FM |
3293 | } else if (key.offset >= off + len) { |
3294 | break; | |
e4355f34 | 3295 | } |
df858e76 | 3296 | next_key_min_offset = key.offset + datal; |
e4355f34 FDBM |
3297 | size = btrfs_item_size_nr(leaf, slot); |
3298 | read_extent_buffer(leaf, buf, | |
3299 | btrfs_item_ptr_offset(leaf, slot), | |
3300 | size); | |
3301 | ||
3302 | btrfs_release_path(path); | |
3303 | path->leave_spinning = 0; | |
c5c9cd4d | 3304 | |
31840ae1 | 3305 | memcpy(&new_key, &key, sizeof(new_key)); |
33345d01 | 3306 | new_key.objectid = btrfs_ino(inode); |
4d728ec7 LZ |
3307 | if (off <= key.offset) |
3308 | new_key.offset = key.offset + destoff - off; | |
3309 | else | |
3310 | new_key.offset = destoff; | |
31840ae1 | 3311 | |
f82a9901 FM |
3312 | /* |
3313 | * Deal with a hole that doesn't have an extent item | |
3314 | * that represents it (NO_HOLES feature enabled). | |
3315 | * This hole is either in the middle of the cloning | |
3316 | * range or at the beginning (fully overlaps it or | |
3317 | * partially overlaps it). | |
3318 | */ | |
3319 | if (new_key.offset != last_dest_end) | |
3320 | drop_start = last_dest_end; | |
3321 | else | |
3322 | drop_start = new_key.offset; | |
3323 | ||
b6f3409b SW |
3324 | /* |
3325 | * 1 - adjusting old extent (we may have to split it) | |
3326 | * 1 - add new extent | |
3327 | * 1 - inode update | |
3328 | */ | |
3329 | trans = btrfs_start_transaction(root, 3); | |
a22285a6 YZ |
3330 | if (IS_ERR(trans)) { |
3331 | ret = PTR_ERR(trans); | |
3332 | goto out; | |
3333 | } | |
3334 | ||
c8a894d7 CM |
3335 | if (type == BTRFS_FILE_EXTENT_REG || |
3336 | type == BTRFS_FILE_EXTENT_PREALLOC) { | |
d72c0842 LZ |
3337 | /* |
3338 | * a | --- range to clone ---| b | |
3339 | * | ------------- extent ------------- | | |
3340 | */ | |
3341 | ||
93915584 | 3342 | /* subtract range b */ |
d72c0842 LZ |
3343 | if (key.offset + datal > off + len) |
3344 | datal = off + len - key.offset; | |
3345 | ||
93915584 | 3346 | /* subtract range a */ |
a22285a6 YZ |
3347 | if (off > key.offset) { |
3348 | datao += off - key.offset; | |
3349 | datal -= off - key.offset; | |
3350 | } | |
3351 | ||
5dc562c5 | 3352 | ret = btrfs_drop_extents(trans, root, inode, |
f82a9901 | 3353 | drop_start, |
a22285a6 | 3354 | new_key.offset + datal, |
2671485d | 3355 | 1); |
79787eaa | 3356 | if (ret) { |
3f9e3df8 | 3357 | if (ret != -EOPNOTSUPP) |
00fdf13a LB |
3358 | btrfs_abort_transaction(trans, |
3359 | root, ret); | |
79787eaa JM |
3360 | btrfs_end_transaction(trans, root); |
3361 | goto out; | |
3362 | } | |
a22285a6 | 3363 | |
c5c9cd4d SW |
3364 | ret = btrfs_insert_empty_item(trans, root, path, |
3365 | &new_key, size); | |
79787eaa JM |
3366 | if (ret) { |
3367 | btrfs_abort_transaction(trans, root, | |
3368 | ret); | |
3369 | btrfs_end_transaction(trans, root); | |
3370 | goto out; | |
3371 | } | |
c5c9cd4d SW |
3372 | |
3373 | leaf = path->nodes[0]; | |
3374 | slot = path->slots[0]; | |
3375 | write_extent_buffer(leaf, buf, | |
31840ae1 ZY |
3376 | btrfs_item_ptr_offset(leaf, slot), |
3377 | size); | |
ae01a0ab | 3378 | |
c5c9cd4d | 3379 | extent = btrfs_item_ptr(leaf, slot, |
f46b5a66 | 3380 | struct btrfs_file_extent_item); |
c5c9cd4d | 3381 | |
c5c9cd4d SW |
3382 | /* disko == 0 means it's a hole */ |
3383 | if (!disko) | |
3384 | datao = 0; | |
c5c9cd4d SW |
3385 | |
3386 | btrfs_set_file_extent_offset(leaf, extent, | |
3387 | datao); | |
3388 | btrfs_set_file_extent_num_bytes(leaf, extent, | |
3389 | datal); | |
fcebe456 JB |
3390 | |
3391 | /* | |
3392 | * We need to look up the roots that point at | |
3393 | * this bytenr and see if the new root does. If | |
3394 | * it does not we need to make sure we update | |
3395 | * quotas appropriately. | |
3396 | */ | |
3397 | if (disko && root != BTRFS_I(src)->root && | |
3398 | disko != last_disko) { | |
3399 | no_quota = check_ref(trans, root, | |
3400 | disko); | |
3401 | if (no_quota < 0) { | |
3402 | btrfs_abort_transaction(trans, | |
3403 | root, | |
3404 | ret); | |
3405 | btrfs_end_transaction(trans, | |
3406 | root); | |
3407 | ret = no_quota; | |
3408 | goto out; | |
3409 | } | |
3410 | } | |
3411 | ||
c5c9cd4d SW |
3412 | if (disko) { |
3413 | inode_add_bytes(inode, datal); | |
ae01a0ab | 3414 | ret = btrfs_inc_extent_ref(trans, root, |
5d4f98a2 YZ |
3415 | disko, diskl, 0, |
3416 | root->root_key.objectid, | |
33345d01 | 3417 | btrfs_ino(inode), |
66d7e7f0 | 3418 | new_key.offset - datao, |
fcebe456 | 3419 | no_quota); |
79787eaa JM |
3420 | if (ret) { |
3421 | btrfs_abort_transaction(trans, | |
3422 | root, | |
3423 | ret); | |
3424 | btrfs_end_transaction(trans, | |
3425 | root); | |
3426 | goto out; | |
3427 | ||
3428 | } | |
f46b5a66 | 3429 | } |
c5c9cd4d SW |
3430 | } else if (type == BTRFS_FILE_EXTENT_INLINE) { |
3431 | u64 skip = 0; | |
3432 | u64 trim = 0; | |
d3ecfcdf LB |
3433 | u64 aligned_end = 0; |
3434 | ||
c5c9cd4d SW |
3435 | if (off > key.offset) { |
3436 | skip = off - key.offset; | |
3437 | new_key.offset += skip; | |
3438 | } | |
d397712b | 3439 | |
aa42ffd9 LB |
3440 | if (key.offset + datal > off + len) |
3441 | trim = key.offset + datal - (off + len); | |
d397712b | 3442 | |
c5c9cd4d | 3443 | if (comp && (skip || trim)) { |
c5c9cd4d | 3444 | ret = -EINVAL; |
a22285a6 | 3445 | btrfs_end_transaction(trans, root); |
c5c9cd4d SW |
3446 | goto out; |
3447 | } | |
3448 | size -= skip + trim; | |
3449 | datal -= skip + trim; | |
a22285a6 | 3450 | |
d3ecfcdf LB |
3451 | aligned_end = ALIGN(new_key.offset + datal, |
3452 | root->sectorsize); | |
5dc562c5 | 3453 | ret = btrfs_drop_extents(trans, root, inode, |
f82a9901 | 3454 | drop_start, |
d3ecfcdf | 3455 | aligned_end, |
2671485d | 3456 | 1); |
79787eaa | 3457 | if (ret) { |
3f9e3df8 | 3458 | if (ret != -EOPNOTSUPP) |
3a29bc09 CM |
3459 | btrfs_abort_transaction(trans, |
3460 | root, ret); | |
79787eaa JM |
3461 | btrfs_end_transaction(trans, root); |
3462 | goto out; | |
3463 | } | |
a22285a6 | 3464 | |
c5c9cd4d SW |
3465 | ret = btrfs_insert_empty_item(trans, root, path, |
3466 | &new_key, size); | |
79787eaa JM |
3467 | if (ret) { |
3468 | btrfs_abort_transaction(trans, root, | |
3469 | ret); | |
3470 | btrfs_end_transaction(trans, root); | |
3471 | goto out; | |
3472 | } | |
c5c9cd4d SW |
3473 | |
3474 | if (skip) { | |
d397712b CM |
3475 | u32 start = |
3476 | btrfs_file_extent_calc_inline_size(0); | |
c5c9cd4d SW |
3477 | memmove(buf+start, buf+start+skip, |
3478 | datal); | |
3479 | } | |
3480 | ||
3481 | leaf = path->nodes[0]; | |
3482 | slot = path->slots[0]; | |
3483 | write_extent_buffer(leaf, buf, | |
3484 | btrfs_item_ptr_offset(leaf, slot), | |
3485 | size); | |
3486 | inode_add_bytes(inode, datal); | |
f46b5a66 | 3487 | } |
c5c9cd4d | 3488 | |
7ffbb598 FM |
3489 | /* If we have an implicit hole (NO_HOLES feature). */ |
3490 | if (drop_start < new_key.offset) | |
3491 | clone_update_extent_map(inode, trans, | |
14f59796 | 3492 | NULL, drop_start, |
7ffbb598 FM |
3493 | new_key.offset - drop_start); |
3494 | ||
14f59796 | 3495 | clone_update_extent_map(inode, trans, path, 0, 0); |
7ffbb598 | 3496 | |
c5c9cd4d | 3497 | btrfs_mark_buffer_dirty(leaf); |
b3b4aa74 | 3498 | btrfs_release_path(path); |
c5c9cd4d | 3499 | |
62e2390e FM |
3500 | last_dest_end = ALIGN(new_key.offset + datal, |
3501 | root->sectorsize); | |
f82a9901 FM |
3502 | ret = clone_finish_inode_update(trans, inode, |
3503 | last_dest_end, | |
3504 | destoff, olen); | |
3505 | if (ret) | |
79787eaa | 3506 | goto out; |
2c463823 FM |
3507 | if (new_key.offset + datal >= destoff + len) |
3508 | break; | |
a22285a6 | 3509 | } |
b3b4aa74 | 3510 | btrfs_release_path(path); |
df858e76 | 3511 | key.offset = next_key_min_offset; |
f46b5a66 | 3512 | } |
f46b5a66 | 3513 | ret = 0; |
32b7c687 | 3514 | |
f82a9901 FM |
3515 | if (last_dest_end < destoff + len) { |
3516 | /* | |
3517 | * We have an implicit hole (NO_HOLES feature is enabled) that | |
3518 | * fully or partially overlaps our cloning range at its end. | |
3519 | */ | |
3520 | btrfs_release_path(path); | |
3521 | ||
3522 | /* | |
3523 | * 1 - remove extent(s) | |
3524 | * 1 - inode update | |
3525 | */ | |
3526 | trans = btrfs_start_transaction(root, 2); | |
3527 | if (IS_ERR(trans)) { | |
3528 | ret = PTR_ERR(trans); | |
3529 | goto out; | |
3530 | } | |
3531 | ret = btrfs_drop_extents(trans, root, inode, | |
3532 | last_dest_end, destoff + len, 1); | |
3533 | if (ret) { | |
3534 | if (ret != -EOPNOTSUPP) | |
3535 | btrfs_abort_transaction(trans, root, ret); | |
3536 | btrfs_end_transaction(trans, root); | |
3537 | goto out; | |
3538 | } | |
14f59796 FM |
3539 | clone_update_extent_map(inode, trans, NULL, last_dest_end, |
3540 | destoff + len - last_dest_end); | |
f82a9901 FM |
3541 | ret = clone_finish_inode_update(trans, inode, destoff + len, |
3542 | destoff, olen); | |
3543 | } | |
3544 | ||
f46b5a66 | 3545 | out: |
32b7c687 MF |
3546 | btrfs_free_path(path); |
3547 | vfree(buf); | |
3548 | return ret; | |
3549 | } | |
3550 | ||
3551 | static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, | |
3552 | u64 off, u64 olen, u64 destoff) | |
3553 | { | |
54563d41 | 3554 | struct inode *inode = file_inode(file); |
32b7c687 MF |
3555 | struct btrfs_root *root = BTRFS_I(inode)->root; |
3556 | struct fd src_file; | |
3557 | struct inode *src; | |
3558 | int ret; | |
3559 | u64 len = olen; | |
3560 | u64 bs = root->fs_info->sb->s_blocksize; | |
3561 | int same_inode = 0; | |
3562 | ||
3563 | /* | |
3564 | * TODO: | |
3565 | * - split compressed inline extents. annoying: we need to | |
3566 | * decompress into destination's address_space (the file offset | |
3567 | * may change, so source mapping won't do), then recompress (or | |
3568 | * otherwise reinsert) a subrange. | |
00fdf13a LB |
3569 | * |
3570 | * - split destination inode's inline extents. The inline extents can | |
3571 | * be either compressed or non-compressed. | |
32b7c687 MF |
3572 | */ |
3573 | ||
3574 | /* the destination must be opened for writing */ | |
3575 | if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) | |
3576 | return -EINVAL; | |
3577 | ||
3578 | if (btrfs_root_readonly(root)) | |
3579 | return -EROFS; | |
3580 | ||
3581 | ret = mnt_want_write_file(file); | |
3582 | if (ret) | |
3583 | return ret; | |
3584 | ||
3585 | src_file = fdget(srcfd); | |
3586 | if (!src_file.file) { | |
3587 | ret = -EBADF; | |
3588 | goto out_drop_write; | |
3589 | } | |
3590 | ||
3591 | ret = -EXDEV; | |
3592 | if (src_file.file->f_path.mnt != file->f_path.mnt) | |
3593 | goto out_fput; | |
3594 | ||
3595 | src = file_inode(src_file.file); | |
3596 | ||
3597 | ret = -EINVAL; | |
3598 | if (src == inode) | |
3599 | same_inode = 1; | |
3600 | ||
3601 | /* the src must be open for reading */ | |
3602 | if (!(src_file.file->f_mode & FMODE_READ)) | |
3603 | goto out_fput; | |
3604 | ||
3605 | /* don't make the dst file partly checksummed */ | |
3606 | if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != | |
3607 | (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) | |
3608 | goto out_fput; | |
3609 | ||
3610 | ret = -EISDIR; | |
3611 | if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) | |
3612 | goto out_fput; | |
3613 | ||
3614 | ret = -EXDEV; | |
3615 | if (src->i_sb != inode->i_sb) | |
3616 | goto out_fput; | |
3617 | ||
3618 | if (!same_inode) { | |
3619 | if (inode < src) { | |
3620 | mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT); | |
3621 | mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD); | |
3622 | } else { | |
3623 | mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT); | |
3624 | mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD); | |
3625 | } | |
3626 | } else { | |
3627 | mutex_lock(&src->i_mutex); | |
3628 | } | |
3629 | ||
3630 | /* determine range to clone */ | |
3631 | ret = -EINVAL; | |
3632 | if (off + len > src->i_size || off + len < off) | |
3633 | goto out_unlock; | |
3634 | if (len == 0) | |
3635 | olen = len = src->i_size - off; | |
3636 | /* if we extend to eof, continue to block boundary */ | |
3637 | if (off + len == src->i_size) | |
3638 | len = ALIGN(src->i_size, bs) - off; | |
3639 | ||
ccccf3d6 FM |
3640 | if (len == 0) { |
3641 | ret = 0; | |
3642 | goto out_unlock; | |
3643 | } | |
3644 | ||
32b7c687 MF |
3645 | /* verify the end result is block aligned */ |
3646 | if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || | |
3647 | !IS_ALIGNED(destoff, bs)) | |
3648 | goto out_unlock; | |
3649 | ||
3650 | /* verify if ranges are overlapped within the same file */ | |
3651 | if (same_inode) { | |
3652 | if (destoff + len > off && destoff < off + len) | |
3653 | goto out_unlock; | |
3654 | } | |
3655 | ||
3656 | if (destoff > inode->i_size) { | |
3657 | ret = btrfs_cont_expand(inode, inode->i_size, destoff); | |
3658 | if (ret) | |
3659 | goto out_unlock; | |
3660 | } | |
3661 | ||
c125b8bf FM |
3662 | /* |
3663 | * Lock the target range too. Right after we replace the file extent | |
3664 | * items in the fs tree (which now point to the cloned data), we might | |
3665 | * have a worker replace them with extent items relative to a write | |
3666 | * operation that was issued before this clone operation (i.e. confront | |
3667 | * with inode.c:btrfs_finish_ordered_io). | |
3668 | */ | |
3669 | if (same_inode) { | |
3670 | u64 lock_start = min_t(u64, off, destoff); | |
3671 | u64 lock_len = max_t(u64, off, destoff) + len - lock_start; | |
32b7c687 | 3672 | |
c125b8bf FM |
3673 | lock_extent_range(src, lock_start, lock_len); |
3674 | } else { | |
3675 | lock_extent_range(src, off, len); | |
3676 | lock_extent_range(inode, destoff, len); | |
3677 | } | |
32b7c687 MF |
3678 | |
3679 | ret = btrfs_clone(src, inode, off, olen, len, destoff); | |
3680 | ||
c125b8bf FM |
3681 | if (same_inode) { |
3682 | u64 lock_start = min_t(u64, off, destoff); | |
3683 | u64 lock_end = max_t(u64, off, destoff) + len - 1; | |
3684 | ||
3685 | unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end); | |
3686 | } else { | |
3687 | unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1); | |
3688 | unlock_extent(&BTRFS_I(inode)->io_tree, destoff, | |
3689 | destoff + len - 1); | |
3690 | } | |
3691 | /* | |
3692 | * Truncate page cache pages so that future reads will see the cloned | |
3693 | * data immediately and not the previous data. | |
3694 | */ | |
3695 | truncate_inode_pages_range(&inode->i_data, destoff, | |
3696 | PAGE_CACHE_ALIGN(destoff + len) - 1); | |
f46b5a66 | 3697 | out_unlock: |
c57c2b3e FDBM |
3698 | if (!same_inode) { |
3699 | if (inode < src) { | |
3700 | mutex_unlock(&src->i_mutex); | |
3701 | mutex_unlock(&inode->i_mutex); | |
3702 | } else { | |
3703 | mutex_unlock(&inode->i_mutex); | |
3704 | mutex_unlock(&src->i_mutex); | |
3705 | } | |
3706 | } else { | |
3707 | mutex_unlock(&src->i_mutex); | |
3708 | } | |
f46b5a66 | 3709 | out_fput: |
2903ff01 | 3710 | fdput(src_file); |
ab67b7c1 | 3711 | out_drop_write: |
2a79f17e | 3712 | mnt_drop_write_file(file); |
f46b5a66 CH |
3713 | return ret; |
3714 | } | |
3715 | ||
7a865e8a | 3716 | static long btrfs_ioctl_clone_range(struct file *file, void __user *argp) |
c5c9cd4d SW |
3717 | { |
3718 | struct btrfs_ioctl_clone_range_args args; | |
3719 | ||
7a865e8a | 3720 | if (copy_from_user(&args, argp, sizeof(args))) |
c5c9cd4d SW |
3721 | return -EFAULT; |
3722 | return btrfs_ioctl_clone(file, args.src_fd, args.src_offset, | |
3723 | args.src_length, args.dest_offset); | |
3724 | } | |
3725 | ||
f46b5a66 CH |
3726 | /* |
3727 | * there are many ways the trans_start and trans_end ioctls can lead | |
3728 | * to deadlocks. They should only be used by applications that | |
3729 | * basically own the machine, and have a very in depth understanding | |
3730 | * of all the possible deadlocks and enospc problems. | |
3731 | */ | |
b2950863 | 3732 | static long btrfs_ioctl_trans_start(struct file *file) |
f46b5a66 | 3733 | { |
496ad9aa | 3734 | struct inode *inode = file_inode(file); |
f46b5a66 CH |
3735 | struct btrfs_root *root = BTRFS_I(inode)->root; |
3736 | struct btrfs_trans_handle *trans; | |
1ab86aed | 3737 | int ret; |
f46b5a66 | 3738 | |
1ab86aed | 3739 | ret = -EPERM; |
df5b5520 | 3740 | if (!capable(CAP_SYS_ADMIN)) |
1ab86aed | 3741 | goto out; |
df5b5520 | 3742 | |
1ab86aed SW |
3743 | ret = -EINPROGRESS; |
3744 | if (file->private_data) | |
f46b5a66 | 3745 | goto out; |
9ca9ee09 | 3746 | |
b83cc969 LZ |
3747 | ret = -EROFS; |
3748 | if (btrfs_root_readonly(root)) | |
3749 | goto out; | |
3750 | ||
a561be71 | 3751 | ret = mnt_want_write_file(file); |
c146afad YZ |
3752 | if (ret) |
3753 | goto out; | |
3754 | ||
a4abeea4 | 3755 | atomic_inc(&root->fs_info->open_ioctl_trans); |
9ca9ee09 | 3756 | |
1ab86aed | 3757 | ret = -ENOMEM; |
7a7eaa40 | 3758 | trans = btrfs_start_ioctl_transaction(root); |
abd30bb0 | 3759 | if (IS_ERR(trans)) |
1ab86aed SW |
3760 | goto out_drop; |
3761 | ||
3762 | file->private_data = trans; | |
3763 | return 0; | |
3764 | ||
3765 | out_drop: | |
a4abeea4 | 3766 | atomic_dec(&root->fs_info->open_ioctl_trans); |
2a79f17e | 3767 | mnt_drop_write_file(file); |
f46b5a66 | 3768 | out: |
f46b5a66 CH |
3769 | return ret; |
3770 | } | |
3771 | ||
6ef5ed0d JB |
3772 | static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) |
3773 | { | |
496ad9aa | 3774 | struct inode *inode = file_inode(file); |
6ef5ed0d JB |
3775 | struct btrfs_root *root = BTRFS_I(inode)->root; |
3776 | struct btrfs_root *new_root; | |
3777 | struct btrfs_dir_item *di; | |
3778 | struct btrfs_trans_handle *trans; | |
3779 | struct btrfs_path *path; | |
3780 | struct btrfs_key location; | |
3781 | struct btrfs_disk_key disk_key; | |
6ef5ed0d JB |
3782 | u64 objectid = 0; |
3783 | u64 dir_id; | |
3c04ce01 | 3784 | int ret; |
6ef5ed0d JB |
3785 | |
3786 | if (!capable(CAP_SYS_ADMIN)) | |
3787 | return -EPERM; | |
3788 | ||
3c04ce01 MX |
3789 | ret = mnt_want_write_file(file); |
3790 | if (ret) | |
3791 | return ret; | |
3792 | ||
3793 | if (copy_from_user(&objectid, argp, sizeof(objectid))) { | |
3794 | ret = -EFAULT; | |
3795 | goto out; | |
3796 | } | |
6ef5ed0d JB |
3797 | |
3798 | if (!objectid) | |
1cecf579 | 3799 | objectid = BTRFS_FS_TREE_OBJECTID; |
6ef5ed0d JB |
3800 | |
3801 | location.objectid = objectid; | |
3802 | location.type = BTRFS_ROOT_ITEM_KEY; | |
3803 | location.offset = (u64)-1; | |
3804 | ||
3805 | new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); | |
3c04ce01 MX |
3806 | if (IS_ERR(new_root)) { |
3807 | ret = PTR_ERR(new_root); | |
3808 | goto out; | |
3809 | } | |
6ef5ed0d | 3810 | |
6ef5ed0d | 3811 | path = btrfs_alloc_path(); |
3c04ce01 MX |
3812 | if (!path) { |
3813 | ret = -ENOMEM; | |
3814 | goto out; | |
3815 | } | |
6ef5ed0d JB |
3816 | path->leave_spinning = 1; |
3817 | ||
3818 | trans = btrfs_start_transaction(root, 1); | |
98d5dc13 | 3819 | if (IS_ERR(trans)) { |
6ef5ed0d | 3820 | btrfs_free_path(path); |
3c04ce01 MX |
3821 | ret = PTR_ERR(trans); |
3822 | goto out; | |
6ef5ed0d JB |
3823 | } |
3824 | ||
6c41761f | 3825 | dir_id = btrfs_super_root_dir(root->fs_info->super_copy); |
6ef5ed0d JB |
3826 | di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path, |
3827 | dir_id, "default", 7, 1); | |
cf1e99a4 | 3828 | if (IS_ERR_OR_NULL(di)) { |
6ef5ed0d JB |
3829 | btrfs_free_path(path); |
3830 | btrfs_end_transaction(trans, root); | |
efe120a0 FH |
3831 | btrfs_err(new_root->fs_info, "Umm, you don't have the default dir" |
3832 | "item, this isn't going to work"); | |
3c04ce01 MX |
3833 | ret = -ENOENT; |
3834 | goto out; | |
6ef5ed0d JB |
3835 | } |
3836 | ||
3837 | btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); | |
3838 | btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); | |
3839 | btrfs_mark_buffer_dirty(path->nodes[0]); | |
3840 | btrfs_free_path(path); | |
3841 | ||
2b0ce2c2 | 3842 | btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL); |
6ef5ed0d | 3843 | btrfs_end_transaction(trans, root); |
3c04ce01 MX |
3844 | out: |
3845 | mnt_drop_write_file(file); | |
3846 | return ret; | |
6ef5ed0d JB |
3847 | } |
3848 | ||
5af3e8cc SB |
3849 | void btrfs_get_block_group_info(struct list_head *groups_list, |
3850 | struct btrfs_ioctl_space_info *space) | |
bf5fc093 JB |
3851 | { |
3852 | struct btrfs_block_group_cache *block_group; | |
3853 | ||
3854 | space->total_bytes = 0; | |
3855 | space->used_bytes = 0; | |
3856 | space->flags = 0; | |
3857 | list_for_each_entry(block_group, groups_list, list) { | |
3858 | space->flags = block_group->flags; | |
3859 | space->total_bytes += block_group->key.offset; | |
3860 | space->used_bytes += | |
3861 | btrfs_block_group_used(&block_group->item); | |
3862 | } | |
3863 | } | |
3864 | ||
48a3b636 | 3865 | static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg) |
1406e432 JB |
3866 | { |
3867 | struct btrfs_ioctl_space_args space_args; | |
3868 | struct btrfs_ioctl_space_info space; | |
3869 | struct btrfs_ioctl_space_info *dest; | |
7fde62bf | 3870 | struct btrfs_ioctl_space_info *dest_orig; |
13f2696f | 3871 | struct btrfs_ioctl_space_info __user *user_dest; |
1406e432 | 3872 | struct btrfs_space_info *info; |
bf5fc093 JB |
3873 | u64 types[] = {BTRFS_BLOCK_GROUP_DATA, |
3874 | BTRFS_BLOCK_GROUP_SYSTEM, | |
3875 | BTRFS_BLOCK_GROUP_METADATA, | |
3876 | BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; | |
3877 | int num_types = 4; | |
7fde62bf | 3878 | int alloc_size; |
1406e432 | 3879 | int ret = 0; |
51788b1b | 3880 | u64 slot_count = 0; |
bf5fc093 | 3881 | int i, c; |
1406e432 JB |
3882 | |
3883 | if (copy_from_user(&space_args, | |
3884 | (struct btrfs_ioctl_space_args __user *)arg, | |
3885 | sizeof(space_args))) | |
3886 | return -EFAULT; | |
3887 | ||
bf5fc093 JB |
3888 | for (i = 0; i < num_types; i++) { |
3889 | struct btrfs_space_info *tmp; | |
3890 | ||
3891 | info = NULL; | |
3892 | rcu_read_lock(); | |
3893 | list_for_each_entry_rcu(tmp, &root->fs_info->space_info, | |
3894 | list) { | |
3895 | if (tmp->flags == types[i]) { | |
3896 | info = tmp; | |
3897 | break; | |
3898 | } | |
3899 | } | |
3900 | rcu_read_unlock(); | |
3901 | ||
3902 | if (!info) | |
3903 | continue; | |
3904 | ||
3905 | down_read(&info->groups_sem); | |
3906 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
3907 | if (!list_empty(&info->block_groups[c])) | |
3908 | slot_count++; | |
3909 | } | |
3910 | up_read(&info->groups_sem); | |
3911 | } | |
7fde62bf | 3912 | |
36523e95 DS |
3913 | /* |
3914 | * Global block reserve, exported as a space_info | |
3915 | */ | |
3916 | slot_count++; | |
3917 | ||
7fde62bf CM |
3918 | /* space_slots == 0 means they are asking for a count */ |
3919 | if (space_args.space_slots == 0) { | |
3920 | space_args.total_spaces = slot_count; | |
3921 | goto out; | |
3922 | } | |
bf5fc093 | 3923 | |
51788b1b | 3924 | slot_count = min_t(u64, space_args.space_slots, slot_count); |
bf5fc093 | 3925 | |
7fde62bf | 3926 | alloc_size = sizeof(*dest) * slot_count; |
bf5fc093 | 3927 | |
7fde62bf CM |
3928 | /* we generally have at most 6 or so space infos, one for each raid |
3929 | * level. So, a whole page should be more than enough for everyone | |
3930 | */ | |
3931 | if (alloc_size > PAGE_CACHE_SIZE) | |
3932 | return -ENOMEM; | |
3933 | ||
1406e432 | 3934 | space_args.total_spaces = 0; |
7fde62bf CM |
3935 | dest = kmalloc(alloc_size, GFP_NOFS); |
3936 | if (!dest) | |
3937 | return -ENOMEM; | |
3938 | dest_orig = dest; | |
1406e432 | 3939 | |
7fde62bf | 3940 | /* now we have a buffer to copy into */ |
bf5fc093 JB |
3941 | for (i = 0; i < num_types; i++) { |
3942 | struct btrfs_space_info *tmp; | |
3943 | ||
51788b1b DR |
3944 | if (!slot_count) |
3945 | break; | |
3946 | ||
bf5fc093 JB |
3947 | info = NULL; |
3948 | rcu_read_lock(); | |
3949 | list_for_each_entry_rcu(tmp, &root->fs_info->space_info, | |
3950 | list) { | |
3951 | if (tmp->flags == types[i]) { | |
3952 | info = tmp; | |
3953 | break; | |
3954 | } | |
3955 | } | |
3956 | rcu_read_unlock(); | |
7fde62bf | 3957 | |
bf5fc093 JB |
3958 | if (!info) |
3959 | continue; | |
3960 | down_read(&info->groups_sem); | |
3961 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
3962 | if (!list_empty(&info->block_groups[c])) { | |
5af3e8cc SB |
3963 | btrfs_get_block_group_info( |
3964 | &info->block_groups[c], &space); | |
bf5fc093 JB |
3965 | memcpy(dest, &space, sizeof(space)); |
3966 | dest++; | |
3967 | space_args.total_spaces++; | |
51788b1b | 3968 | slot_count--; |
bf5fc093 | 3969 | } |
51788b1b DR |
3970 | if (!slot_count) |
3971 | break; | |
bf5fc093 JB |
3972 | } |
3973 | up_read(&info->groups_sem); | |
1406e432 | 3974 | } |
1406e432 | 3975 | |
36523e95 DS |
3976 | /* |
3977 | * Add global block reserve | |
3978 | */ | |
3979 | if (slot_count) { | |
3980 | struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv; | |
3981 | ||
3982 | spin_lock(&block_rsv->lock); | |
3983 | space.total_bytes = block_rsv->size; | |
3984 | space.used_bytes = block_rsv->size - block_rsv->reserved; | |
3985 | spin_unlock(&block_rsv->lock); | |
3986 | space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; | |
3987 | memcpy(dest, &space, sizeof(space)); | |
3988 | space_args.total_spaces++; | |
3989 | } | |
3990 | ||
2eec6c81 | 3991 | user_dest = (struct btrfs_ioctl_space_info __user *) |
7fde62bf CM |
3992 | (arg + sizeof(struct btrfs_ioctl_space_args)); |
3993 | ||
3994 | if (copy_to_user(user_dest, dest_orig, alloc_size)) | |
3995 | ret = -EFAULT; | |
3996 | ||
3997 | kfree(dest_orig); | |
3998 | out: | |
3999 | if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) | |
1406e432 JB |
4000 | ret = -EFAULT; |
4001 | ||
4002 | return ret; | |
4003 | } | |
4004 | ||
f46b5a66 CH |
4005 | /* |
4006 | * there are many ways the trans_start and trans_end ioctls can lead | |
4007 | * to deadlocks. They should only be used by applications that | |
4008 | * basically own the machine, and have a very in depth understanding | |
4009 | * of all the possible deadlocks and enospc problems. | |
4010 | */ | |
4011 | long btrfs_ioctl_trans_end(struct file *file) | |
4012 | { | |
496ad9aa | 4013 | struct inode *inode = file_inode(file); |
f46b5a66 CH |
4014 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4015 | struct btrfs_trans_handle *trans; | |
f46b5a66 | 4016 | |
f46b5a66 | 4017 | trans = file->private_data; |
1ab86aed SW |
4018 | if (!trans) |
4019 | return -EINVAL; | |
b214107e | 4020 | file->private_data = NULL; |
9ca9ee09 | 4021 | |
1ab86aed SW |
4022 | btrfs_end_transaction(trans, root); |
4023 | ||
a4abeea4 | 4024 | atomic_dec(&root->fs_info->open_ioctl_trans); |
9ca9ee09 | 4025 | |
2a79f17e | 4026 | mnt_drop_write_file(file); |
1ab86aed | 4027 | return 0; |
f46b5a66 CH |
4028 | } |
4029 | ||
9a8c28be MX |
4030 | static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, |
4031 | void __user *argp) | |
46204592 | 4032 | { |
46204592 SW |
4033 | struct btrfs_trans_handle *trans; |
4034 | u64 transid; | |
db5b493a | 4035 | int ret; |
46204592 | 4036 | |
d4edf39b | 4037 | trans = btrfs_attach_transaction_barrier(root); |
ff7c1d33 MX |
4038 | if (IS_ERR(trans)) { |
4039 | if (PTR_ERR(trans) != -ENOENT) | |
4040 | return PTR_ERR(trans); | |
4041 | ||
4042 | /* No running transaction, don't bother */ | |
4043 | transid = root->fs_info->last_trans_committed; | |
4044 | goto out; | |
4045 | } | |
46204592 | 4046 | transid = trans->transid; |
db5b493a | 4047 | ret = btrfs_commit_transaction_async(trans, root, 0); |
8b2b2d3c TI |
4048 | if (ret) { |
4049 | btrfs_end_transaction(trans, root); | |
db5b493a | 4050 | return ret; |
8b2b2d3c | 4051 | } |
ff7c1d33 | 4052 | out: |
46204592 SW |
4053 | if (argp) |
4054 | if (copy_to_user(argp, &transid, sizeof(transid))) | |
4055 | return -EFAULT; | |
4056 | return 0; | |
4057 | } | |
4058 | ||
9a8c28be MX |
4059 | static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root, |
4060 | void __user *argp) | |
46204592 | 4061 | { |
46204592 SW |
4062 | u64 transid; |
4063 | ||
4064 | if (argp) { | |
4065 | if (copy_from_user(&transid, argp, sizeof(transid))) | |
4066 | return -EFAULT; | |
4067 | } else { | |
4068 | transid = 0; /* current trans */ | |
4069 | } | |
4070 | return btrfs_wait_for_commit(root, transid); | |
4071 | } | |
4072 | ||
b8e95489 | 4073 | static long btrfs_ioctl_scrub(struct file *file, void __user *arg) |
475f6387 | 4074 | { |
496ad9aa | 4075 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
475f6387 | 4076 | struct btrfs_ioctl_scrub_args *sa; |
b8e95489 | 4077 | int ret; |
475f6387 JS |
4078 | |
4079 | if (!capable(CAP_SYS_ADMIN)) | |
4080 | return -EPERM; | |
4081 | ||
4082 | sa = memdup_user(arg, sizeof(*sa)); | |
4083 | if (IS_ERR(sa)) | |
4084 | return PTR_ERR(sa); | |
4085 | ||
b8e95489 MX |
4086 | if (!(sa->flags & BTRFS_SCRUB_READONLY)) { |
4087 | ret = mnt_want_write_file(file); | |
4088 | if (ret) | |
4089 | goto out; | |
4090 | } | |
4091 | ||
aa1b8cd4 | 4092 | ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end, |
63a212ab SB |
4093 | &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, |
4094 | 0); | |
475f6387 JS |
4095 | |
4096 | if (copy_to_user(arg, sa, sizeof(*sa))) | |
4097 | ret = -EFAULT; | |
4098 | ||
b8e95489 MX |
4099 | if (!(sa->flags & BTRFS_SCRUB_READONLY)) |
4100 | mnt_drop_write_file(file); | |
4101 | out: | |
475f6387 JS |
4102 | kfree(sa); |
4103 | return ret; | |
4104 | } | |
4105 | ||
4106 | static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg) | |
4107 | { | |
4108 | if (!capable(CAP_SYS_ADMIN)) | |
4109 | return -EPERM; | |
4110 | ||
aa1b8cd4 | 4111 | return btrfs_scrub_cancel(root->fs_info); |
475f6387 JS |
4112 | } |
4113 | ||
4114 | static long btrfs_ioctl_scrub_progress(struct btrfs_root *root, | |
4115 | void __user *arg) | |
4116 | { | |
4117 | struct btrfs_ioctl_scrub_args *sa; | |
4118 | int ret; | |
4119 | ||
4120 | if (!capable(CAP_SYS_ADMIN)) | |
4121 | return -EPERM; | |
4122 | ||
4123 | sa = memdup_user(arg, sizeof(*sa)); | |
4124 | if (IS_ERR(sa)) | |
4125 | return PTR_ERR(sa); | |
4126 | ||
4127 | ret = btrfs_scrub_progress(root, sa->devid, &sa->progress); | |
4128 | ||
4129 | if (copy_to_user(arg, sa, sizeof(*sa))) | |
4130 | ret = -EFAULT; | |
4131 | ||
4132 | kfree(sa); | |
4133 | return ret; | |
4134 | } | |
4135 | ||
c11d2c23 | 4136 | static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root, |
b27f7c0c | 4137 | void __user *arg) |
c11d2c23 SB |
4138 | { |
4139 | struct btrfs_ioctl_get_dev_stats *sa; | |
4140 | int ret; | |
4141 | ||
c11d2c23 SB |
4142 | sa = memdup_user(arg, sizeof(*sa)); |
4143 | if (IS_ERR(sa)) | |
4144 | return PTR_ERR(sa); | |
4145 | ||
b27f7c0c DS |
4146 | if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { |
4147 | kfree(sa); | |
4148 | return -EPERM; | |
4149 | } | |
4150 | ||
4151 | ret = btrfs_get_dev_stats(root, sa); | |
c11d2c23 SB |
4152 | |
4153 | if (copy_to_user(arg, sa, sizeof(*sa))) | |
4154 | ret = -EFAULT; | |
4155 | ||
4156 | kfree(sa); | |
4157 | return ret; | |
4158 | } | |
4159 | ||
3f6bcfbd SB |
4160 | static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg) |
4161 | { | |
4162 | struct btrfs_ioctl_dev_replace_args *p; | |
4163 | int ret; | |
4164 | ||
4165 | if (!capable(CAP_SYS_ADMIN)) | |
4166 | return -EPERM; | |
4167 | ||
4168 | p = memdup_user(arg, sizeof(*p)); | |
4169 | if (IS_ERR(p)) | |
4170 | return PTR_ERR(p); | |
4171 | ||
4172 | switch (p->cmd) { | |
4173 | case BTRFS_IOCTL_DEV_REPLACE_CMD_START: | |
adfa97cb ID |
4174 | if (root->fs_info->sb->s_flags & MS_RDONLY) { |
4175 | ret = -EROFS; | |
4176 | goto out; | |
4177 | } | |
3f6bcfbd SB |
4178 | if (atomic_xchg( |
4179 | &root->fs_info->mutually_exclusive_operation_running, | |
4180 | 1)) { | |
e57138b3 | 4181 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
3f6bcfbd SB |
4182 | } else { |
4183 | ret = btrfs_dev_replace_start(root, p); | |
4184 | atomic_set( | |
4185 | &root->fs_info->mutually_exclusive_operation_running, | |
4186 | 0); | |
4187 | } | |
4188 | break; | |
4189 | case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: | |
4190 | btrfs_dev_replace_status(root->fs_info, p); | |
4191 | ret = 0; | |
4192 | break; | |
4193 | case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: | |
4194 | ret = btrfs_dev_replace_cancel(root->fs_info, p); | |
4195 | break; | |
4196 | default: | |
4197 | ret = -EINVAL; | |
4198 | break; | |
4199 | } | |
4200 | ||
4201 | if (copy_to_user(arg, p, sizeof(*p))) | |
4202 | ret = -EFAULT; | |
adfa97cb | 4203 | out: |
3f6bcfbd SB |
4204 | kfree(p); |
4205 | return ret; | |
4206 | } | |
4207 | ||
d7728c96 JS |
4208 | static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) |
4209 | { | |
4210 | int ret = 0; | |
4211 | int i; | |
740c3d22 | 4212 | u64 rel_ptr; |
d7728c96 | 4213 | int size; |
806468f8 | 4214 | struct btrfs_ioctl_ino_path_args *ipa = NULL; |
d7728c96 JS |
4215 | struct inode_fs_paths *ipath = NULL; |
4216 | struct btrfs_path *path; | |
4217 | ||
82b22ac8 | 4218 | if (!capable(CAP_DAC_READ_SEARCH)) |
d7728c96 JS |
4219 | return -EPERM; |
4220 | ||
4221 | path = btrfs_alloc_path(); | |
4222 | if (!path) { | |
4223 | ret = -ENOMEM; | |
4224 | goto out; | |
4225 | } | |
4226 | ||
4227 | ipa = memdup_user(arg, sizeof(*ipa)); | |
4228 | if (IS_ERR(ipa)) { | |
4229 | ret = PTR_ERR(ipa); | |
4230 | ipa = NULL; | |
4231 | goto out; | |
4232 | } | |
4233 | ||
4234 | size = min_t(u32, ipa->size, 4096); | |
4235 | ipath = init_ipath(size, root, path); | |
4236 | if (IS_ERR(ipath)) { | |
4237 | ret = PTR_ERR(ipath); | |
4238 | ipath = NULL; | |
4239 | goto out; | |
4240 | } | |
4241 | ||
4242 | ret = paths_from_inode(ipa->inum, ipath); | |
4243 | if (ret < 0) | |
4244 | goto out; | |
4245 | ||
4246 | for (i = 0; i < ipath->fspath->elem_cnt; ++i) { | |
745c4d8e JM |
4247 | rel_ptr = ipath->fspath->val[i] - |
4248 | (u64)(unsigned long)ipath->fspath->val; | |
740c3d22 | 4249 | ipath->fspath->val[i] = rel_ptr; |
d7728c96 JS |
4250 | } |
4251 | ||
745c4d8e JM |
4252 | ret = copy_to_user((void *)(unsigned long)ipa->fspath, |
4253 | (void *)(unsigned long)ipath->fspath, size); | |
d7728c96 JS |
4254 | if (ret) { |
4255 | ret = -EFAULT; | |
4256 | goto out; | |
4257 | } | |
4258 | ||
4259 | out: | |
4260 | btrfs_free_path(path); | |
4261 | free_ipath(ipath); | |
4262 | kfree(ipa); | |
4263 | ||
4264 | return ret; | |
4265 | } | |
4266 | ||
4267 | static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) | |
4268 | { | |
4269 | struct btrfs_data_container *inodes = ctx; | |
4270 | const size_t c = 3 * sizeof(u64); | |
4271 | ||
4272 | if (inodes->bytes_left >= c) { | |
4273 | inodes->bytes_left -= c; | |
4274 | inodes->val[inodes->elem_cnt] = inum; | |
4275 | inodes->val[inodes->elem_cnt + 1] = offset; | |
4276 | inodes->val[inodes->elem_cnt + 2] = root; | |
4277 | inodes->elem_cnt += 3; | |
4278 | } else { | |
4279 | inodes->bytes_missing += c - inodes->bytes_left; | |
4280 | inodes->bytes_left = 0; | |
4281 | inodes->elem_missed += 3; | |
4282 | } | |
4283 | ||
4284 | return 0; | |
4285 | } | |
4286 | ||
4287 | static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root, | |
4288 | void __user *arg) | |
4289 | { | |
4290 | int ret = 0; | |
4291 | int size; | |
d7728c96 JS |
4292 | struct btrfs_ioctl_logical_ino_args *loi; |
4293 | struct btrfs_data_container *inodes = NULL; | |
4294 | struct btrfs_path *path = NULL; | |
d7728c96 JS |
4295 | |
4296 | if (!capable(CAP_SYS_ADMIN)) | |
4297 | return -EPERM; | |
4298 | ||
4299 | loi = memdup_user(arg, sizeof(*loi)); | |
4300 | if (IS_ERR(loi)) { | |
4301 | ret = PTR_ERR(loi); | |
4302 | loi = NULL; | |
4303 | goto out; | |
4304 | } | |
4305 | ||
4306 | path = btrfs_alloc_path(); | |
4307 | if (!path) { | |
4308 | ret = -ENOMEM; | |
4309 | goto out; | |
4310 | } | |
4311 | ||
425d17a2 | 4312 | size = min_t(u32, loi->size, 64 * 1024); |
d7728c96 JS |
4313 | inodes = init_data_container(size); |
4314 | if (IS_ERR(inodes)) { | |
4315 | ret = PTR_ERR(inodes); | |
4316 | inodes = NULL; | |
4317 | goto out; | |
4318 | } | |
4319 | ||
df031f07 LB |
4320 | ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path, |
4321 | build_ino_list, inodes); | |
4322 | if (ret == -EINVAL) | |
d7728c96 JS |
4323 | ret = -ENOENT; |
4324 | if (ret < 0) | |
4325 | goto out; | |
4326 | ||
745c4d8e JM |
4327 | ret = copy_to_user((void *)(unsigned long)loi->inodes, |
4328 | (void *)(unsigned long)inodes, size); | |
d7728c96 JS |
4329 | if (ret) |
4330 | ret = -EFAULT; | |
4331 | ||
4332 | out: | |
4333 | btrfs_free_path(path); | |
425d17a2 | 4334 | vfree(inodes); |
d7728c96 JS |
4335 | kfree(loi); |
4336 | ||
4337 | return ret; | |
4338 | } | |
4339 | ||
19a39dce | 4340 | void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock, |
c9e9f97b ID |
4341 | struct btrfs_ioctl_balance_args *bargs) |
4342 | { | |
4343 | struct btrfs_balance_control *bctl = fs_info->balance_ctl; | |
4344 | ||
4345 | bargs->flags = bctl->flags; | |
4346 | ||
837d5b6e ID |
4347 | if (atomic_read(&fs_info->balance_running)) |
4348 | bargs->state |= BTRFS_BALANCE_STATE_RUNNING; | |
4349 | if (atomic_read(&fs_info->balance_pause_req)) | |
4350 | bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; | |
a7e99c69 ID |
4351 | if (atomic_read(&fs_info->balance_cancel_req)) |
4352 | bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; | |
837d5b6e | 4353 | |
c9e9f97b ID |
4354 | memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); |
4355 | memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); | |
4356 | memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); | |
19a39dce ID |
4357 | |
4358 | if (lock) { | |
4359 | spin_lock(&fs_info->balance_lock); | |
4360 | memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); | |
4361 | spin_unlock(&fs_info->balance_lock); | |
4362 | } else { | |
4363 | memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); | |
4364 | } | |
c9e9f97b ID |
4365 | } |
4366 | ||
9ba1f6e4 | 4367 | static long btrfs_ioctl_balance(struct file *file, void __user *arg) |
c9e9f97b | 4368 | { |
496ad9aa | 4369 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
c9e9f97b ID |
4370 | struct btrfs_fs_info *fs_info = root->fs_info; |
4371 | struct btrfs_ioctl_balance_args *bargs; | |
4372 | struct btrfs_balance_control *bctl; | |
ed0fb78f | 4373 | bool need_unlock; /* for mut. excl. ops lock */ |
c9e9f97b ID |
4374 | int ret; |
4375 | ||
4376 | if (!capable(CAP_SYS_ADMIN)) | |
4377 | return -EPERM; | |
4378 | ||
e54bfa31 | 4379 | ret = mnt_want_write_file(file); |
9ba1f6e4 LB |
4380 | if (ret) |
4381 | return ret; | |
4382 | ||
ed0fb78f ID |
4383 | again: |
4384 | if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { | |
4385 | mutex_lock(&fs_info->volume_mutex); | |
4386 | mutex_lock(&fs_info->balance_mutex); | |
4387 | need_unlock = true; | |
4388 | goto locked; | |
4389 | } | |
4390 | ||
4391 | /* | |
4392 | * mut. excl. ops lock is locked. Three possibilites: | |
4393 | * (1) some other op is running | |
4394 | * (2) balance is running | |
4395 | * (3) balance is paused -- special case (think resume) | |
4396 | */ | |
c9e9f97b | 4397 | mutex_lock(&fs_info->balance_mutex); |
ed0fb78f ID |
4398 | if (fs_info->balance_ctl) { |
4399 | /* this is either (2) or (3) */ | |
4400 | if (!atomic_read(&fs_info->balance_running)) { | |
4401 | mutex_unlock(&fs_info->balance_mutex); | |
4402 | if (!mutex_trylock(&fs_info->volume_mutex)) | |
4403 | goto again; | |
4404 | mutex_lock(&fs_info->balance_mutex); | |
4405 | ||
4406 | if (fs_info->balance_ctl && | |
4407 | !atomic_read(&fs_info->balance_running)) { | |
4408 | /* this is (3) */ | |
4409 | need_unlock = false; | |
4410 | goto locked; | |
4411 | } | |
4412 | ||
4413 | mutex_unlock(&fs_info->balance_mutex); | |
4414 | mutex_unlock(&fs_info->volume_mutex); | |
4415 | goto again; | |
4416 | } else { | |
4417 | /* this is (2) */ | |
4418 | mutex_unlock(&fs_info->balance_mutex); | |
4419 | ret = -EINPROGRESS; | |
4420 | goto out; | |
4421 | } | |
4422 | } else { | |
4423 | /* this is (1) */ | |
4424 | mutex_unlock(&fs_info->balance_mutex); | |
e57138b3 | 4425 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
ed0fb78f ID |
4426 | goto out; |
4427 | } | |
4428 | ||
4429 | locked: | |
4430 | BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running)); | |
c9e9f97b ID |
4431 | |
4432 | if (arg) { | |
4433 | bargs = memdup_user(arg, sizeof(*bargs)); | |
4434 | if (IS_ERR(bargs)) { | |
4435 | ret = PTR_ERR(bargs); | |
ed0fb78f | 4436 | goto out_unlock; |
c9e9f97b | 4437 | } |
de322263 ID |
4438 | |
4439 | if (bargs->flags & BTRFS_BALANCE_RESUME) { | |
4440 | if (!fs_info->balance_ctl) { | |
4441 | ret = -ENOTCONN; | |
4442 | goto out_bargs; | |
4443 | } | |
4444 | ||
4445 | bctl = fs_info->balance_ctl; | |
4446 | spin_lock(&fs_info->balance_lock); | |
4447 | bctl->flags |= BTRFS_BALANCE_RESUME; | |
4448 | spin_unlock(&fs_info->balance_lock); | |
4449 | ||
4450 | goto do_balance; | |
4451 | } | |
c9e9f97b ID |
4452 | } else { |
4453 | bargs = NULL; | |
4454 | } | |
4455 | ||
ed0fb78f | 4456 | if (fs_info->balance_ctl) { |
837d5b6e ID |
4457 | ret = -EINPROGRESS; |
4458 | goto out_bargs; | |
4459 | } | |
4460 | ||
c9e9f97b ID |
4461 | bctl = kzalloc(sizeof(*bctl), GFP_NOFS); |
4462 | if (!bctl) { | |
4463 | ret = -ENOMEM; | |
4464 | goto out_bargs; | |
4465 | } | |
4466 | ||
4467 | bctl->fs_info = fs_info; | |
4468 | if (arg) { | |
4469 | memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); | |
4470 | memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); | |
4471 | memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); | |
4472 | ||
4473 | bctl->flags = bargs->flags; | |
f43ffb60 ID |
4474 | } else { |
4475 | /* balance everything - no filters */ | |
4476 | bctl->flags |= BTRFS_BALANCE_TYPE_MASK; | |
c9e9f97b ID |
4477 | } |
4478 | ||
de322263 | 4479 | do_balance: |
c9e9f97b | 4480 | /* |
ed0fb78f ID |
4481 | * Ownership of bctl and mutually_exclusive_operation_running |
4482 | * goes to to btrfs_balance. bctl is freed in __cancel_balance, | |
4483 | * or, if restriper was paused all the way until unmount, in | |
4484 | * free_fs_info. mutually_exclusive_operation_running is | |
4485 | * cleared in __cancel_balance. | |
c9e9f97b | 4486 | */ |
ed0fb78f ID |
4487 | need_unlock = false; |
4488 | ||
4489 | ret = btrfs_balance(bctl, bargs); | |
4490 | ||
c9e9f97b ID |
4491 | if (arg) { |
4492 | if (copy_to_user(arg, bargs, sizeof(*bargs))) | |
4493 | ret = -EFAULT; | |
4494 | } | |
4495 | ||
4496 | out_bargs: | |
4497 | kfree(bargs); | |
ed0fb78f | 4498 | out_unlock: |
c9e9f97b ID |
4499 | mutex_unlock(&fs_info->balance_mutex); |
4500 | mutex_unlock(&fs_info->volume_mutex); | |
ed0fb78f ID |
4501 | if (need_unlock) |
4502 | atomic_set(&fs_info->mutually_exclusive_operation_running, 0); | |
4503 | out: | |
e54bfa31 | 4504 | mnt_drop_write_file(file); |
c9e9f97b ID |
4505 | return ret; |
4506 | } | |
4507 | ||
837d5b6e ID |
4508 | static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd) |
4509 | { | |
4510 | if (!capable(CAP_SYS_ADMIN)) | |
4511 | return -EPERM; | |
4512 | ||
4513 | switch (cmd) { | |
4514 | case BTRFS_BALANCE_CTL_PAUSE: | |
4515 | return btrfs_pause_balance(root->fs_info); | |
a7e99c69 ID |
4516 | case BTRFS_BALANCE_CTL_CANCEL: |
4517 | return btrfs_cancel_balance(root->fs_info); | |
837d5b6e ID |
4518 | } |
4519 | ||
4520 | return -EINVAL; | |
4521 | } | |
4522 | ||
19a39dce ID |
4523 | static long btrfs_ioctl_balance_progress(struct btrfs_root *root, |
4524 | void __user *arg) | |
4525 | { | |
4526 | struct btrfs_fs_info *fs_info = root->fs_info; | |
4527 | struct btrfs_ioctl_balance_args *bargs; | |
4528 | int ret = 0; | |
4529 | ||
4530 | if (!capable(CAP_SYS_ADMIN)) | |
4531 | return -EPERM; | |
4532 | ||
4533 | mutex_lock(&fs_info->balance_mutex); | |
4534 | if (!fs_info->balance_ctl) { | |
4535 | ret = -ENOTCONN; | |
4536 | goto out; | |
4537 | } | |
4538 | ||
4539 | bargs = kzalloc(sizeof(*bargs), GFP_NOFS); | |
4540 | if (!bargs) { | |
4541 | ret = -ENOMEM; | |
4542 | goto out; | |
4543 | } | |
4544 | ||
4545 | update_ioctl_balance_args(fs_info, 1, bargs); | |
4546 | ||
4547 | if (copy_to_user(arg, bargs, sizeof(*bargs))) | |
4548 | ret = -EFAULT; | |
4549 | ||
4550 | kfree(bargs); | |
4551 | out: | |
4552 | mutex_unlock(&fs_info->balance_mutex); | |
4553 | return ret; | |
4554 | } | |
4555 | ||
905b0dda | 4556 | static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) |
5d13a37b | 4557 | { |
496ad9aa | 4558 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
5d13a37b AJ |
4559 | struct btrfs_ioctl_quota_ctl_args *sa; |
4560 | struct btrfs_trans_handle *trans = NULL; | |
4561 | int ret; | |
4562 | int err; | |
4563 | ||
4564 | if (!capable(CAP_SYS_ADMIN)) | |
4565 | return -EPERM; | |
4566 | ||
905b0dda MX |
4567 | ret = mnt_want_write_file(file); |
4568 | if (ret) | |
4569 | return ret; | |
5d13a37b AJ |
4570 | |
4571 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4572 | if (IS_ERR(sa)) { |
4573 | ret = PTR_ERR(sa); | |
4574 | goto drop_write; | |
4575 | } | |
5d13a37b | 4576 | |
7708f029 | 4577 | down_write(&root->fs_info->subvol_sem); |
2f232036 JS |
4578 | trans = btrfs_start_transaction(root->fs_info->tree_root, 2); |
4579 | if (IS_ERR(trans)) { | |
4580 | ret = PTR_ERR(trans); | |
4581 | goto out; | |
5d13a37b AJ |
4582 | } |
4583 | ||
4584 | switch (sa->cmd) { | |
4585 | case BTRFS_QUOTA_CTL_ENABLE: | |
4586 | ret = btrfs_quota_enable(trans, root->fs_info); | |
4587 | break; | |
4588 | case BTRFS_QUOTA_CTL_DISABLE: | |
4589 | ret = btrfs_quota_disable(trans, root->fs_info); | |
4590 | break; | |
5d13a37b AJ |
4591 | default: |
4592 | ret = -EINVAL; | |
4593 | break; | |
4594 | } | |
4595 | ||
2f232036 JS |
4596 | err = btrfs_commit_transaction(trans, root->fs_info->tree_root); |
4597 | if (err && !ret) | |
4598 | ret = err; | |
5d13a37b AJ |
4599 | out: |
4600 | kfree(sa); | |
7708f029 | 4601 | up_write(&root->fs_info->subvol_sem); |
905b0dda MX |
4602 | drop_write: |
4603 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4604 | return ret; |
4605 | } | |
4606 | ||
905b0dda | 4607 | static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) |
5d13a37b | 4608 | { |
496ad9aa | 4609 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
5d13a37b AJ |
4610 | struct btrfs_ioctl_qgroup_assign_args *sa; |
4611 | struct btrfs_trans_handle *trans; | |
4612 | int ret; | |
4613 | int err; | |
4614 | ||
4615 | if (!capable(CAP_SYS_ADMIN)) | |
4616 | return -EPERM; | |
4617 | ||
905b0dda MX |
4618 | ret = mnt_want_write_file(file); |
4619 | if (ret) | |
4620 | return ret; | |
5d13a37b AJ |
4621 | |
4622 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4623 | if (IS_ERR(sa)) { |
4624 | ret = PTR_ERR(sa); | |
4625 | goto drop_write; | |
4626 | } | |
5d13a37b AJ |
4627 | |
4628 | trans = btrfs_join_transaction(root); | |
4629 | if (IS_ERR(trans)) { | |
4630 | ret = PTR_ERR(trans); | |
4631 | goto out; | |
4632 | } | |
4633 | ||
4634 | /* FIXME: check if the IDs really exist */ | |
4635 | if (sa->assign) { | |
4636 | ret = btrfs_add_qgroup_relation(trans, root->fs_info, | |
4637 | sa->src, sa->dst); | |
4638 | } else { | |
4639 | ret = btrfs_del_qgroup_relation(trans, root->fs_info, | |
4640 | sa->src, sa->dst); | |
4641 | } | |
4642 | ||
e082f563 QW |
4643 | /* update qgroup status and info */ |
4644 | err = btrfs_run_qgroups(trans, root->fs_info); | |
4645 | if (err < 0) | |
4646 | btrfs_error(root->fs_info, ret, | |
4647 | "failed to update qgroup status and info\n"); | |
5d13a37b AJ |
4648 | err = btrfs_end_transaction(trans, root); |
4649 | if (err && !ret) | |
4650 | ret = err; | |
4651 | ||
4652 | out: | |
4653 | kfree(sa); | |
905b0dda MX |
4654 | drop_write: |
4655 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4656 | return ret; |
4657 | } | |
4658 | ||
905b0dda | 4659 | static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) |
5d13a37b | 4660 | { |
496ad9aa | 4661 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
5d13a37b AJ |
4662 | struct btrfs_ioctl_qgroup_create_args *sa; |
4663 | struct btrfs_trans_handle *trans; | |
4664 | int ret; | |
4665 | int err; | |
4666 | ||
4667 | if (!capable(CAP_SYS_ADMIN)) | |
4668 | return -EPERM; | |
4669 | ||
905b0dda MX |
4670 | ret = mnt_want_write_file(file); |
4671 | if (ret) | |
4672 | return ret; | |
5d13a37b AJ |
4673 | |
4674 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4675 | if (IS_ERR(sa)) { |
4676 | ret = PTR_ERR(sa); | |
4677 | goto drop_write; | |
4678 | } | |
5d13a37b | 4679 | |
d86e56cf MX |
4680 | if (!sa->qgroupid) { |
4681 | ret = -EINVAL; | |
4682 | goto out; | |
4683 | } | |
4684 | ||
5d13a37b AJ |
4685 | trans = btrfs_join_transaction(root); |
4686 | if (IS_ERR(trans)) { | |
4687 | ret = PTR_ERR(trans); | |
4688 | goto out; | |
4689 | } | |
4690 | ||
4691 | /* FIXME: check if the IDs really exist */ | |
4692 | if (sa->create) { | |
4087cf24 | 4693 | ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid); |
5d13a37b AJ |
4694 | } else { |
4695 | ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid); | |
4696 | } | |
4697 | ||
4698 | err = btrfs_end_transaction(trans, root); | |
4699 | if (err && !ret) | |
4700 | ret = err; | |
4701 | ||
4702 | out: | |
4703 | kfree(sa); | |
905b0dda MX |
4704 | drop_write: |
4705 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4706 | return ret; |
4707 | } | |
4708 | ||
905b0dda | 4709 | static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) |
5d13a37b | 4710 | { |
496ad9aa | 4711 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
5d13a37b AJ |
4712 | struct btrfs_ioctl_qgroup_limit_args *sa; |
4713 | struct btrfs_trans_handle *trans; | |
4714 | int ret; | |
4715 | int err; | |
4716 | u64 qgroupid; | |
4717 | ||
4718 | if (!capable(CAP_SYS_ADMIN)) | |
4719 | return -EPERM; | |
4720 | ||
905b0dda MX |
4721 | ret = mnt_want_write_file(file); |
4722 | if (ret) | |
4723 | return ret; | |
5d13a37b AJ |
4724 | |
4725 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4726 | if (IS_ERR(sa)) { |
4727 | ret = PTR_ERR(sa); | |
4728 | goto drop_write; | |
4729 | } | |
5d13a37b AJ |
4730 | |
4731 | trans = btrfs_join_transaction(root); | |
4732 | if (IS_ERR(trans)) { | |
4733 | ret = PTR_ERR(trans); | |
4734 | goto out; | |
4735 | } | |
4736 | ||
4737 | qgroupid = sa->qgroupid; | |
4738 | if (!qgroupid) { | |
4739 | /* take the current subvol as qgroup */ | |
4740 | qgroupid = root->root_key.objectid; | |
4741 | } | |
4742 | ||
4743 | /* FIXME: check if the IDs really exist */ | |
4744 | ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim); | |
4745 | ||
4746 | err = btrfs_end_transaction(trans, root); | |
4747 | if (err && !ret) | |
4748 | ret = err; | |
4749 | ||
4750 | out: | |
4751 | kfree(sa); | |
905b0dda MX |
4752 | drop_write: |
4753 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4754 | return ret; |
4755 | } | |
4756 | ||
2f232036 JS |
4757 | static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) |
4758 | { | |
6d0379ec | 4759 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
2f232036 JS |
4760 | struct btrfs_ioctl_quota_rescan_args *qsa; |
4761 | int ret; | |
4762 | ||
4763 | if (!capable(CAP_SYS_ADMIN)) | |
4764 | return -EPERM; | |
4765 | ||
4766 | ret = mnt_want_write_file(file); | |
4767 | if (ret) | |
4768 | return ret; | |
4769 | ||
4770 | qsa = memdup_user(arg, sizeof(*qsa)); | |
4771 | if (IS_ERR(qsa)) { | |
4772 | ret = PTR_ERR(qsa); | |
4773 | goto drop_write; | |
4774 | } | |
4775 | ||
4776 | if (qsa->flags) { | |
4777 | ret = -EINVAL; | |
4778 | goto out; | |
4779 | } | |
4780 | ||
4781 | ret = btrfs_qgroup_rescan(root->fs_info); | |
4782 | ||
4783 | out: | |
4784 | kfree(qsa); | |
4785 | drop_write: | |
4786 | mnt_drop_write_file(file); | |
4787 | return ret; | |
4788 | } | |
4789 | ||
4790 | static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) | |
4791 | { | |
6d0379ec | 4792 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
2f232036 JS |
4793 | struct btrfs_ioctl_quota_rescan_args *qsa; |
4794 | int ret = 0; | |
4795 | ||
4796 | if (!capable(CAP_SYS_ADMIN)) | |
4797 | return -EPERM; | |
4798 | ||
4799 | qsa = kzalloc(sizeof(*qsa), GFP_NOFS); | |
4800 | if (!qsa) | |
4801 | return -ENOMEM; | |
4802 | ||
4803 | if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { | |
4804 | qsa->flags = 1; | |
4805 | qsa->progress = root->fs_info->qgroup_rescan_progress.objectid; | |
4806 | } | |
4807 | ||
4808 | if (copy_to_user(arg, qsa, sizeof(*qsa))) | |
4809 | ret = -EFAULT; | |
4810 | ||
4811 | kfree(qsa); | |
4812 | return ret; | |
4813 | } | |
4814 | ||
57254b6e JS |
4815 | static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) |
4816 | { | |
54563d41 | 4817 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
57254b6e JS |
4818 | |
4819 | if (!capable(CAP_SYS_ADMIN)) | |
4820 | return -EPERM; | |
4821 | ||
4822 | return btrfs_qgroup_wait_for_completion(root->fs_info); | |
4823 | } | |
4824 | ||
abccd00f HM |
4825 | static long _btrfs_ioctl_set_received_subvol(struct file *file, |
4826 | struct btrfs_ioctl_received_subvol_args *sa) | |
8ea05e3a | 4827 | { |
496ad9aa | 4828 | struct inode *inode = file_inode(file); |
8ea05e3a AB |
4829 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4830 | struct btrfs_root_item *root_item = &root->root_item; | |
4831 | struct btrfs_trans_handle *trans; | |
4832 | struct timespec ct = CURRENT_TIME; | |
4833 | int ret = 0; | |
dd5f9615 | 4834 | int received_uuid_changed; |
8ea05e3a | 4835 | |
bd60ea0f DS |
4836 | if (!inode_owner_or_capable(inode)) |
4837 | return -EPERM; | |
4838 | ||
8ea05e3a AB |
4839 | ret = mnt_want_write_file(file); |
4840 | if (ret < 0) | |
4841 | return ret; | |
4842 | ||
4843 | down_write(&root->fs_info->subvol_sem); | |
4844 | ||
4845 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { | |
4846 | ret = -EINVAL; | |
4847 | goto out; | |
4848 | } | |
4849 | ||
4850 | if (btrfs_root_readonly(root)) { | |
4851 | ret = -EROFS; | |
4852 | goto out; | |
4853 | } | |
4854 | ||
dd5f9615 SB |
4855 | /* |
4856 | * 1 - root item | |
4857 | * 2 - uuid items (received uuid + subvol uuid) | |
4858 | */ | |
4859 | trans = btrfs_start_transaction(root, 3); | |
8ea05e3a AB |
4860 | if (IS_ERR(trans)) { |
4861 | ret = PTR_ERR(trans); | |
4862 | trans = NULL; | |
4863 | goto out; | |
4864 | } | |
4865 | ||
4866 | sa->rtransid = trans->transid; | |
4867 | sa->rtime.sec = ct.tv_sec; | |
4868 | sa->rtime.nsec = ct.tv_nsec; | |
4869 | ||
dd5f9615 SB |
4870 | received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, |
4871 | BTRFS_UUID_SIZE); | |
4872 | if (received_uuid_changed && | |
4873 | !btrfs_is_empty_uuid(root_item->received_uuid)) | |
4874 | btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, | |
4875 | root_item->received_uuid, | |
4876 | BTRFS_UUID_KEY_RECEIVED_SUBVOL, | |
4877 | root->root_key.objectid); | |
8ea05e3a AB |
4878 | memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); |
4879 | btrfs_set_root_stransid(root_item, sa->stransid); | |
4880 | btrfs_set_root_rtransid(root_item, sa->rtransid); | |
3cae210f QW |
4881 | btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); |
4882 | btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); | |
4883 | btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); | |
4884 | btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); | |
8ea05e3a AB |
4885 | |
4886 | ret = btrfs_update_root(trans, root->fs_info->tree_root, | |
4887 | &root->root_key, &root->root_item); | |
4888 | if (ret < 0) { | |
4889 | btrfs_end_transaction(trans, root); | |
8ea05e3a | 4890 | goto out; |
dd5f9615 SB |
4891 | } |
4892 | if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { | |
4893 | ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root, | |
4894 | sa->uuid, | |
4895 | BTRFS_UUID_KEY_RECEIVED_SUBVOL, | |
4896 | root->root_key.objectid); | |
4897 | if (ret < 0 && ret != -EEXIST) { | |
4898 | btrfs_abort_transaction(trans, root, ret); | |
8ea05e3a | 4899 | goto out; |
dd5f9615 SB |
4900 | } |
4901 | } | |
4902 | ret = btrfs_commit_transaction(trans, root); | |
4903 | if (ret < 0) { | |
4904 | btrfs_abort_transaction(trans, root, ret); | |
4905 | goto out; | |
8ea05e3a AB |
4906 | } |
4907 | ||
abccd00f HM |
4908 | out: |
4909 | up_write(&root->fs_info->subvol_sem); | |
4910 | mnt_drop_write_file(file); | |
4911 | return ret; | |
4912 | } | |
4913 | ||
4914 | #ifdef CONFIG_64BIT | |
4915 | static long btrfs_ioctl_set_received_subvol_32(struct file *file, | |
4916 | void __user *arg) | |
4917 | { | |
4918 | struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; | |
4919 | struct btrfs_ioctl_received_subvol_args *args64 = NULL; | |
4920 | int ret = 0; | |
4921 | ||
4922 | args32 = memdup_user(arg, sizeof(*args32)); | |
4923 | if (IS_ERR(args32)) { | |
4924 | ret = PTR_ERR(args32); | |
4925 | args32 = NULL; | |
4926 | goto out; | |
4927 | } | |
4928 | ||
4929 | args64 = kmalloc(sizeof(*args64), GFP_NOFS); | |
84dbeb87 DC |
4930 | if (!args64) { |
4931 | ret = -ENOMEM; | |
abccd00f HM |
4932 | goto out; |
4933 | } | |
4934 | ||
4935 | memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); | |
4936 | args64->stransid = args32->stransid; | |
4937 | args64->rtransid = args32->rtransid; | |
4938 | args64->stime.sec = args32->stime.sec; | |
4939 | args64->stime.nsec = args32->stime.nsec; | |
4940 | args64->rtime.sec = args32->rtime.sec; | |
4941 | args64->rtime.nsec = args32->rtime.nsec; | |
4942 | args64->flags = args32->flags; | |
4943 | ||
4944 | ret = _btrfs_ioctl_set_received_subvol(file, args64); | |
4945 | if (ret) | |
4946 | goto out; | |
4947 | ||
4948 | memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); | |
4949 | args32->stransid = args64->stransid; | |
4950 | args32->rtransid = args64->rtransid; | |
4951 | args32->stime.sec = args64->stime.sec; | |
4952 | args32->stime.nsec = args64->stime.nsec; | |
4953 | args32->rtime.sec = args64->rtime.sec; | |
4954 | args32->rtime.nsec = args64->rtime.nsec; | |
4955 | args32->flags = args64->flags; | |
4956 | ||
4957 | ret = copy_to_user(arg, args32, sizeof(*args32)); | |
4958 | if (ret) | |
4959 | ret = -EFAULT; | |
4960 | ||
4961 | out: | |
4962 | kfree(args32); | |
4963 | kfree(args64); | |
4964 | return ret; | |
4965 | } | |
4966 | #endif | |
4967 | ||
4968 | static long btrfs_ioctl_set_received_subvol(struct file *file, | |
4969 | void __user *arg) | |
4970 | { | |
4971 | struct btrfs_ioctl_received_subvol_args *sa = NULL; | |
4972 | int ret = 0; | |
4973 | ||
4974 | sa = memdup_user(arg, sizeof(*sa)); | |
4975 | if (IS_ERR(sa)) { | |
4976 | ret = PTR_ERR(sa); | |
4977 | sa = NULL; | |
4978 | goto out; | |
4979 | } | |
4980 | ||
4981 | ret = _btrfs_ioctl_set_received_subvol(file, sa); | |
4982 | ||
4983 | if (ret) | |
4984 | goto out; | |
4985 | ||
8ea05e3a AB |
4986 | ret = copy_to_user(arg, sa, sizeof(*sa)); |
4987 | if (ret) | |
4988 | ret = -EFAULT; | |
4989 | ||
4990 | out: | |
4991 | kfree(sa); | |
8ea05e3a AB |
4992 | return ret; |
4993 | } | |
4994 | ||
867ab667 | 4995 | static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) |
4996 | { | |
6d0379ec | 4997 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
a1b83ac5 | 4998 | size_t len; |
867ab667 | 4999 | int ret; |
a1b83ac5 AJ |
5000 | char label[BTRFS_LABEL_SIZE]; |
5001 | ||
5002 | spin_lock(&root->fs_info->super_lock); | |
5003 | memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE); | |
5004 | spin_unlock(&root->fs_info->super_lock); | |
5005 | ||
5006 | len = strnlen(label, BTRFS_LABEL_SIZE); | |
867ab667 | 5007 | |
5008 | if (len == BTRFS_LABEL_SIZE) { | |
efe120a0 FH |
5009 | btrfs_warn(root->fs_info, |
5010 | "label is too long, return the first %zu bytes", --len); | |
867ab667 | 5011 | } |
5012 | ||
867ab667 | 5013 | ret = copy_to_user(arg, label, len); |
867ab667 | 5014 | |
5015 | return ret ? -EFAULT : 0; | |
5016 | } | |
5017 | ||
a8bfd4ab | 5018 | static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) |
5019 | { | |
6d0379ec | 5020 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
a8bfd4ab | 5021 | struct btrfs_super_block *super_block = root->fs_info->super_copy; |
5022 | struct btrfs_trans_handle *trans; | |
5023 | char label[BTRFS_LABEL_SIZE]; | |
5024 | int ret; | |
5025 | ||
5026 | if (!capable(CAP_SYS_ADMIN)) | |
5027 | return -EPERM; | |
5028 | ||
5029 | if (copy_from_user(label, arg, sizeof(label))) | |
5030 | return -EFAULT; | |
5031 | ||
5032 | if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { | |
efe120a0 | 5033 | btrfs_err(root->fs_info, "unable to set label with more than %d bytes", |
a8bfd4ab | 5034 | BTRFS_LABEL_SIZE - 1); |
5035 | return -EINVAL; | |
5036 | } | |
5037 | ||
5038 | ret = mnt_want_write_file(file); | |
5039 | if (ret) | |
5040 | return ret; | |
5041 | ||
a8bfd4ab | 5042 | trans = btrfs_start_transaction(root, 0); |
5043 | if (IS_ERR(trans)) { | |
5044 | ret = PTR_ERR(trans); | |
5045 | goto out_unlock; | |
5046 | } | |
5047 | ||
a1b83ac5 | 5048 | spin_lock(&root->fs_info->super_lock); |
a8bfd4ab | 5049 | strcpy(super_block->label, label); |
a1b83ac5 | 5050 | spin_unlock(&root->fs_info->super_lock); |
d0270aca | 5051 | ret = btrfs_commit_transaction(trans, root); |
a8bfd4ab | 5052 | |
5053 | out_unlock: | |
a8bfd4ab | 5054 | mnt_drop_write_file(file); |
5055 | return ret; | |
5056 | } | |
5057 | ||
2eaa055f JM |
5058 | #define INIT_FEATURE_FLAGS(suffix) \ |
5059 | { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ | |
5060 | .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ | |
5061 | .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } | |
5062 | ||
5063 | static int btrfs_ioctl_get_supported_features(struct file *file, | |
5064 | void __user *arg) | |
5065 | { | |
5066 | static struct btrfs_ioctl_feature_flags features[3] = { | |
5067 | INIT_FEATURE_FLAGS(SUPP), | |
5068 | INIT_FEATURE_FLAGS(SAFE_SET), | |
5069 | INIT_FEATURE_FLAGS(SAFE_CLEAR) | |
5070 | }; | |
5071 | ||
5072 | if (copy_to_user(arg, &features, sizeof(features))) | |
5073 | return -EFAULT; | |
5074 | ||
5075 | return 0; | |
5076 | } | |
5077 | ||
5078 | static int btrfs_ioctl_get_features(struct file *file, void __user *arg) | |
5079 | { | |
5080 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; | |
5081 | struct btrfs_super_block *super_block = root->fs_info->super_copy; | |
5082 | struct btrfs_ioctl_feature_flags features; | |
5083 | ||
5084 | features.compat_flags = btrfs_super_compat_flags(super_block); | |
5085 | features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); | |
5086 | features.incompat_flags = btrfs_super_incompat_flags(super_block); | |
5087 | ||
5088 | if (copy_to_user(arg, &features, sizeof(features))) | |
5089 | return -EFAULT; | |
5090 | ||
5091 | return 0; | |
5092 | } | |
5093 | ||
3b02a68a JM |
5094 | static int check_feature_bits(struct btrfs_root *root, |
5095 | enum btrfs_feature_set set, | |
2eaa055f JM |
5096 | u64 change_mask, u64 flags, u64 supported_flags, |
5097 | u64 safe_set, u64 safe_clear) | |
5098 | { | |
3b02a68a JM |
5099 | const char *type = btrfs_feature_set_names[set]; |
5100 | char *names; | |
2eaa055f JM |
5101 | u64 disallowed, unsupported; |
5102 | u64 set_mask = flags & change_mask; | |
5103 | u64 clear_mask = ~flags & change_mask; | |
5104 | ||
5105 | unsupported = set_mask & ~supported_flags; | |
5106 | if (unsupported) { | |
3b02a68a JM |
5107 | names = btrfs_printable_features(set, unsupported); |
5108 | if (names) { | |
5109 | btrfs_warn(root->fs_info, | |
5110 | "this kernel does not support the %s feature bit%s", | |
5111 | names, strchr(names, ',') ? "s" : ""); | |
5112 | kfree(names); | |
5113 | } else | |
5114 | btrfs_warn(root->fs_info, | |
2eaa055f JM |
5115 | "this kernel does not support %s bits 0x%llx", |
5116 | type, unsupported); | |
5117 | return -EOPNOTSUPP; | |
5118 | } | |
5119 | ||
5120 | disallowed = set_mask & ~safe_set; | |
5121 | if (disallowed) { | |
3b02a68a JM |
5122 | names = btrfs_printable_features(set, disallowed); |
5123 | if (names) { | |
5124 | btrfs_warn(root->fs_info, | |
5125 | "can't set the %s feature bit%s while mounted", | |
5126 | names, strchr(names, ',') ? "s" : ""); | |
5127 | kfree(names); | |
5128 | } else | |
5129 | btrfs_warn(root->fs_info, | |
2eaa055f JM |
5130 | "can't set %s bits 0x%llx while mounted", |
5131 | type, disallowed); | |
5132 | return -EPERM; | |
5133 | } | |
5134 | ||
5135 | disallowed = clear_mask & ~safe_clear; | |
5136 | if (disallowed) { | |
3b02a68a JM |
5137 | names = btrfs_printable_features(set, disallowed); |
5138 | if (names) { | |
5139 | btrfs_warn(root->fs_info, | |
5140 | "can't clear the %s feature bit%s while mounted", | |
5141 | names, strchr(names, ',') ? "s" : ""); | |
5142 | kfree(names); | |
5143 | } else | |
5144 | btrfs_warn(root->fs_info, | |
2eaa055f JM |
5145 | "can't clear %s bits 0x%llx while mounted", |
5146 | type, disallowed); | |
5147 | return -EPERM; | |
5148 | } | |
5149 | ||
5150 | return 0; | |
5151 | } | |
5152 | ||
5153 | #define check_feature(root, change_mask, flags, mask_base) \ | |
3b02a68a | 5154 | check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \ |
2eaa055f JM |
5155 | BTRFS_FEATURE_ ## mask_base ## _SUPP, \ |
5156 | BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ | |
5157 | BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) | |
5158 | ||
5159 | static int btrfs_ioctl_set_features(struct file *file, void __user *arg) | |
5160 | { | |
5161 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; | |
5162 | struct btrfs_super_block *super_block = root->fs_info->super_copy; | |
5163 | struct btrfs_ioctl_feature_flags flags[2]; | |
5164 | struct btrfs_trans_handle *trans; | |
5165 | u64 newflags; | |
5166 | int ret; | |
5167 | ||
5168 | if (!capable(CAP_SYS_ADMIN)) | |
5169 | return -EPERM; | |
5170 | ||
5171 | if (copy_from_user(flags, arg, sizeof(flags))) | |
5172 | return -EFAULT; | |
5173 | ||
5174 | /* Nothing to do */ | |
5175 | if (!flags[0].compat_flags && !flags[0].compat_ro_flags && | |
5176 | !flags[0].incompat_flags) | |
5177 | return 0; | |
5178 | ||
5179 | ret = check_feature(root, flags[0].compat_flags, | |
5180 | flags[1].compat_flags, COMPAT); | |
5181 | if (ret) | |
5182 | return ret; | |
5183 | ||
5184 | ret = check_feature(root, flags[0].compat_ro_flags, | |
5185 | flags[1].compat_ro_flags, COMPAT_RO); | |
5186 | if (ret) | |
5187 | return ret; | |
5188 | ||
5189 | ret = check_feature(root, flags[0].incompat_flags, | |
5190 | flags[1].incompat_flags, INCOMPAT); | |
5191 | if (ret) | |
5192 | return ret; | |
5193 | ||
8051aa1a | 5194 | trans = btrfs_start_transaction(root, 0); |
2eaa055f JM |
5195 | if (IS_ERR(trans)) |
5196 | return PTR_ERR(trans); | |
5197 | ||
5198 | spin_lock(&root->fs_info->super_lock); | |
5199 | newflags = btrfs_super_compat_flags(super_block); | |
5200 | newflags |= flags[0].compat_flags & flags[1].compat_flags; | |
5201 | newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); | |
5202 | btrfs_set_super_compat_flags(super_block, newflags); | |
5203 | ||
5204 | newflags = btrfs_super_compat_ro_flags(super_block); | |
5205 | newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; | |
5206 | newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); | |
5207 | btrfs_set_super_compat_ro_flags(super_block, newflags); | |
5208 | ||
5209 | newflags = btrfs_super_incompat_flags(super_block); | |
5210 | newflags |= flags[0].incompat_flags & flags[1].incompat_flags; | |
5211 | newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); | |
5212 | btrfs_set_super_incompat_flags(super_block, newflags); | |
5213 | spin_unlock(&root->fs_info->super_lock); | |
5214 | ||
d0270aca | 5215 | return btrfs_commit_transaction(trans, root); |
2eaa055f JM |
5216 | } |
5217 | ||
f46b5a66 CH |
5218 | long btrfs_ioctl(struct file *file, unsigned int |
5219 | cmd, unsigned long arg) | |
5220 | { | |
496ad9aa | 5221 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
4bcabaa3 | 5222 | void __user *argp = (void __user *)arg; |
f46b5a66 CH |
5223 | |
5224 | switch (cmd) { | |
6cbff00f CH |
5225 | case FS_IOC_GETFLAGS: |
5226 | return btrfs_ioctl_getflags(file, argp); | |
5227 | case FS_IOC_SETFLAGS: | |
5228 | return btrfs_ioctl_setflags(file, argp); | |
5229 | case FS_IOC_GETVERSION: | |
5230 | return btrfs_ioctl_getversion(file, argp); | |
f7039b1d LD |
5231 | case FITRIM: |
5232 | return btrfs_ioctl_fitrim(file, argp); | |
f46b5a66 | 5233 | case BTRFS_IOC_SNAP_CREATE: |
fa0d2b9b | 5234 | return btrfs_ioctl_snap_create(file, argp, 0); |
fdfb1e4f | 5235 | case BTRFS_IOC_SNAP_CREATE_V2: |
fa0d2b9b | 5236 | return btrfs_ioctl_snap_create_v2(file, argp, 0); |
3de4586c | 5237 | case BTRFS_IOC_SUBVOL_CREATE: |
fa0d2b9b | 5238 | return btrfs_ioctl_snap_create(file, argp, 1); |
6f72c7e2 AJ |
5239 | case BTRFS_IOC_SUBVOL_CREATE_V2: |
5240 | return btrfs_ioctl_snap_create_v2(file, argp, 1); | |
76dda93c YZ |
5241 | case BTRFS_IOC_SNAP_DESTROY: |
5242 | return btrfs_ioctl_snap_destroy(file, argp); | |
0caa102d LZ |
5243 | case BTRFS_IOC_SUBVOL_GETFLAGS: |
5244 | return btrfs_ioctl_subvol_getflags(file, argp); | |
5245 | case BTRFS_IOC_SUBVOL_SETFLAGS: | |
5246 | return btrfs_ioctl_subvol_setflags(file, argp); | |
6ef5ed0d JB |
5247 | case BTRFS_IOC_DEFAULT_SUBVOL: |
5248 | return btrfs_ioctl_default_subvol(file, argp); | |
f46b5a66 | 5249 | case BTRFS_IOC_DEFRAG: |
1e701a32 CM |
5250 | return btrfs_ioctl_defrag(file, NULL); |
5251 | case BTRFS_IOC_DEFRAG_RANGE: | |
5252 | return btrfs_ioctl_defrag(file, argp); | |
f46b5a66 | 5253 | case BTRFS_IOC_RESIZE: |
198605a8 | 5254 | return btrfs_ioctl_resize(file, argp); |
f46b5a66 | 5255 | case BTRFS_IOC_ADD_DEV: |
4bcabaa3 | 5256 | return btrfs_ioctl_add_dev(root, argp); |
f46b5a66 | 5257 | case BTRFS_IOC_RM_DEV: |
da24927b | 5258 | return btrfs_ioctl_rm_dev(file, argp); |
475f6387 JS |
5259 | case BTRFS_IOC_FS_INFO: |
5260 | return btrfs_ioctl_fs_info(root, argp); | |
5261 | case BTRFS_IOC_DEV_INFO: | |
5262 | return btrfs_ioctl_dev_info(root, argp); | |
f46b5a66 | 5263 | case BTRFS_IOC_BALANCE: |
9ba1f6e4 | 5264 | return btrfs_ioctl_balance(file, NULL); |
f46b5a66 | 5265 | case BTRFS_IOC_CLONE: |
c5c9cd4d SW |
5266 | return btrfs_ioctl_clone(file, arg, 0, 0, 0); |
5267 | case BTRFS_IOC_CLONE_RANGE: | |
7a865e8a | 5268 | return btrfs_ioctl_clone_range(file, argp); |
f46b5a66 CH |
5269 | case BTRFS_IOC_TRANS_START: |
5270 | return btrfs_ioctl_trans_start(file); | |
5271 | case BTRFS_IOC_TRANS_END: | |
5272 | return btrfs_ioctl_trans_end(file); | |
ac8e9819 CM |
5273 | case BTRFS_IOC_TREE_SEARCH: |
5274 | return btrfs_ioctl_tree_search(file, argp); | |
cc68a8a5 GH |
5275 | case BTRFS_IOC_TREE_SEARCH_V2: |
5276 | return btrfs_ioctl_tree_search_v2(file, argp); | |
ac8e9819 CM |
5277 | case BTRFS_IOC_INO_LOOKUP: |
5278 | return btrfs_ioctl_ino_lookup(file, argp); | |
d7728c96 JS |
5279 | case BTRFS_IOC_INO_PATHS: |
5280 | return btrfs_ioctl_ino_to_path(root, argp); | |
5281 | case BTRFS_IOC_LOGICAL_INO: | |
5282 | return btrfs_ioctl_logical_to_ino(root, argp); | |
1406e432 JB |
5283 | case BTRFS_IOC_SPACE_INFO: |
5284 | return btrfs_ioctl_space_info(root, argp); | |
9b199859 FDBM |
5285 | case BTRFS_IOC_SYNC: { |
5286 | int ret; | |
5287 | ||
6c255e67 | 5288 | ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1); |
9b199859 FDBM |
5289 | if (ret) |
5290 | return ret; | |
ddb52f4f | 5291 | ret = btrfs_sync_fs(file_inode(file)->i_sb, 1); |
2fad4e83 DS |
5292 | /* |
5293 | * The transaction thread may want to do more work, | |
5294 | * namely it pokes the cleaner ktread that will start | |
5295 | * processing uncleaned subvols. | |
5296 | */ | |
5297 | wake_up_process(root->fs_info->transaction_kthread); | |
9b199859 FDBM |
5298 | return ret; |
5299 | } | |
46204592 | 5300 | case BTRFS_IOC_START_SYNC: |
9a8c28be | 5301 | return btrfs_ioctl_start_sync(root, argp); |
46204592 | 5302 | case BTRFS_IOC_WAIT_SYNC: |
9a8c28be | 5303 | return btrfs_ioctl_wait_sync(root, argp); |
475f6387 | 5304 | case BTRFS_IOC_SCRUB: |
b8e95489 | 5305 | return btrfs_ioctl_scrub(file, argp); |
475f6387 JS |
5306 | case BTRFS_IOC_SCRUB_CANCEL: |
5307 | return btrfs_ioctl_scrub_cancel(root, argp); | |
5308 | case BTRFS_IOC_SCRUB_PROGRESS: | |
5309 | return btrfs_ioctl_scrub_progress(root, argp); | |
c9e9f97b | 5310 | case BTRFS_IOC_BALANCE_V2: |
9ba1f6e4 | 5311 | return btrfs_ioctl_balance(file, argp); |
837d5b6e ID |
5312 | case BTRFS_IOC_BALANCE_CTL: |
5313 | return btrfs_ioctl_balance_ctl(root, arg); | |
19a39dce ID |
5314 | case BTRFS_IOC_BALANCE_PROGRESS: |
5315 | return btrfs_ioctl_balance_progress(root, argp); | |
8ea05e3a AB |
5316 | case BTRFS_IOC_SET_RECEIVED_SUBVOL: |
5317 | return btrfs_ioctl_set_received_subvol(file, argp); | |
abccd00f HM |
5318 | #ifdef CONFIG_64BIT |
5319 | case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: | |
5320 | return btrfs_ioctl_set_received_subvol_32(file, argp); | |
5321 | #endif | |
31db9f7c AB |
5322 | case BTRFS_IOC_SEND: |
5323 | return btrfs_ioctl_send(file, argp); | |
c11d2c23 | 5324 | case BTRFS_IOC_GET_DEV_STATS: |
b27f7c0c | 5325 | return btrfs_ioctl_get_dev_stats(root, argp); |
5d13a37b | 5326 | case BTRFS_IOC_QUOTA_CTL: |
905b0dda | 5327 | return btrfs_ioctl_quota_ctl(file, argp); |
5d13a37b | 5328 | case BTRFS_IOC_QGROUP_ASSIGN: |
905b0dda | 5329 | return btrfs_ioctl_qgroup_assign(file, argp); |
5d13a37b | 5330 | case BTRFS_IOC_QGROUP_CREATE: |
905b0dda | 5331 | return btrfs_ioctl_qgroup_create(file, argp); |
5d13a37b | 5332 | case BTRFS_IOC_QGROUP_LIMIT: |
905b0dda | 5333 | return btrfs_ioctl_qgroup_limit(file, argp); |
2f232036 JS |
5334 | case BTRFS_IOC_QUOTA_RESCAN: |
5335 | return btrfs_ioctl_quota_rescan(file, argp); | |
5336 | case BTRFS_IOC_QUOTA_RESCAN_STATUS: | |
5337 | return btrfs_ioctl_quota_rescan_status(file, argp); | |
57254b6e JS |
5338 | case BTRFS_IOC_QUOTA_RESCAN_WAIT: |
5339 | return btrfs_ioctl_quota_rescan_wait(file, argp); | |
3f6bcfbd SB |
5340 | case BTRFS_IOC_DEV_REPLACE: |
5341 | return btrfs_ioctl_dev_replace(root, argp); | |
867ab667 | 5342 | case BTRFS_IOC_GET_FSLABEL: |
5343 | return btrfs_ioctl_get_fslabel(file, argp); | |
a8bfd4ab | 5344 | case BTRFS_IOC_SET_FSLABEL: |
5345 | return btrfs_ioctl_set_fslabel(file, argp); | |
416161db MF |
5346 | case BTRFS_IOC_FILE_EXTENT_SAME: |
5347 | return btrfs_ioctl_file_extent_same(file, argp); | |
2eaa055f JM |
5348 | case BTRFS_IOC_GET_SUPPORTED_FEATURES: |
5349 | return btrfs_ioctl_get_supported_features(file, argp); | |
5350 | case BTRFS_IOC_GET_FEATURES: | |
5351 | return btrfs_ioctl_get_features(file, argp); | |
5352 | case BTRFS_IOC_SET_FEATURES: | |
5353 | return btrfs_ioctl_set_features(file, argp); | |
f46b5a66 CH |
5354 | } |
5355 | ||
5356 | return -ENOTTY; | |
5357 | } |