]>
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" |
f46b5a66 | 61 | |
abccd00f HM |
62 | #ifdef CONFIG_64BIT |
63 | /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI | |
64 | * structures are incorrect, as the timespec structure from userspace | |
65 | * is 4 bytes too small. We define these alternatives here to teach | |
66 | * the kernel about the 32-bit struct packing. | |
67 | */ | |
68 | struct btrfs_ioctl_timespec_32 { | |
69 | __u64 sec; | |
70 | __u32 nsec; | |
71 | } __attribute__ ((__packed__)); | |
72 | ||
73 | struct btrfs_ioctl_received_subvol_args_32 { | |
74 | char uuid[BTRFS_UUID_SIZE]; /* in */ | |
75 | __u64 stransid; /* in */ | |
76 | __u64 rtransid; /* out */ | |
77 | struct btrfs_ioctl_timespec_32 stime; /* in */ | |
78 | struct btrfs_ioctl_timespec_32 rtime; /* out */ | |
79 | __u64 flags; /* in */ | |
80 | __u64 reserved[16]; /* in */ | |
81 | } __attribute__ ((__packed__)); | |
82 | ||
83 | #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ | |
84 | struct btrfs_ioctl_received_subvol_args_32) | |
85 | #endif | |
86 | ||
87 | ||
416161db MF |
88 | static int btrfs_clone(struct inode *src, struct inode *inode, |
89 | u64 off, u64 olen, u64 olen_aligned, u64 destoff); | |
90 | ||
6cbff00f CH |
91 | /* Mask out flags that are inappropriate for the given type of inode. */ |
92 | static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags) | |
93 | { | |
94 | if (S_ISDIR(mode)) | |
95 | return flags; | |
96 | else if (S_ISREG(mode)) | |
97 | return flags & ~FS_DIRSYNC_FL; | |
98 | else | |
99 | return flags & (FS_NODUMP_FL | FS_NOATIME_FL); | |
100 | } | |
101 | ||
102 | /* | |
103 | * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl. | |
104 | */ | |
105 | static unsigned int btrfs_flags_to_ioctl(unsigned int flags) | |
106 | { | |
107 | unsigned int iflags = 0; | |
108 | ||
109 | if (flags & BTRFS_INODE_SYNC) | |
110 | iflags |= FS_SYNC_FL; | |
111 | if (flags & BTRFS_INODE_IMMUTABLE) | |
112 | iflags |= FS_IMMUTABLE_FL; | |
113 | if (flags & BTRFS_INODE_APPEND) | |
114 | iflags |= FS_APPEND_FL; | |
115 | if (flags & BTRFS_INODE_NODUMP) | |
116 | iflags |= FS_NODUMP_FL; | |
117 | if (flags & BTRFS_INODE_NOATIME) | |
118 | iflags |= FS_NOATIME_FL; | |
119 | if (flags & BTRFS_INODE_DIRSYNC) | |
120 | iflags |= FS_DIRSYNC_FL; | |
d0092bdd LZ |
121 | if (flags & BTRFS_INODE_NODATACOW) |
122 | iflags |= FS_NOCOW_FL; | |
123 | ||
124 | if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS)) | |
125 | iflags |= FS_COMPR_FL; | |
126 | else if (flags & BTRFS_INODE_NOCOMPRESS) | |
127 | iflags |= FS_NOCOMP_FL; | |
6cbff00f CH |
128 | |
129 | return iflags; | |
130 | } | |
131 | ||
132 | /* | |
133 | * Update inode->i_flags based on the btrfs internal flags. | |
134 | */ | |
135 | void btrfs_update_iflags(struct inode *inode) | |
136 | { | |
137 | struct btrfs_inode *ip = BTRFS_I(inode); | |
138 | ||
139 | inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC); | |
140 | ||
141 | if (ip->flags & BTRFS_INODE_SYNC) | |
142 | inode->i_flags |= S_SYNC; | |
143 | if (ip->flags & BTRFS_INODE_IMMUTABLE) | |
144 | inode->i_flags |= S_IMMUTABLE; | |
145 | if (ip->flags & BTRFS_INODE_APPEND) | |
146 | inode->i_flags |= S_APPEND; | |
147 | if (ip->flags & BTRFS_INODE_NOATIME) | |
148 | inode->i_flags |= S_NOATIME; | |
149 | if (ip->flags & BTRFS_INODE_DIRSYNC) | |
150 | inode->i_flags |= S_DIRSYNC; | |
151 | } | |
152 | ||
153 | /* | |
154 | * Inherit flags from the parent inode. | |
155 | * | |
e27425d6 | 156 | * Currently only the compression flags and the cow flags are inherited. |
6cbff00f CH |
157 | */ |
158 | void btrfs_inherit_iflags(struct inode *inode, struct inode *dir) | |
159 | { | |
0b4dcea5 CM |
160 | unsigned int flags; |
161 | ||
162 | if (!dir) | |
163 | return; | |
164 | ||
165 | flags = BTRFS_I(dir)->flags; | |
6cbff00f | 166 | |
e27425d6 JB |
167 | if (flags & BTRFS_INODE_NOCOMPRESS) { |
168 | BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS; | |
169 | BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS; | |
170 | } else if (flags & BTRFS_INODE_COMPRESS) { | |
171 | BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS; | |
172 | BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS; | |
173 | } | |
174 | ||
213490b3 | 175 | if (flags & BTRFS_INODE_NODATACOW) { |
e27425d6 | 176 | BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW; |
213490b3 LB |
177 | if (S_ISREG(inode->i_mode)) |
178 | BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM; | |
179 | } | |
6cbff00f | 180 | |
6cbff00f CH |
181 | btrfs_update_iflags(inode); |
182 | } | |
183 | ||
184 | static int btrfs_ioctl_getflags(struct file *file, void __user *arg) | |
185 | { | |
496ad9aa | 186 | struct btrfs_inode *ip = BTRFS_I(file_inode(file)); |
6cbff00f CH |
187 | unsigned int flags = btrfs_flags_to_ioctl(ip->flags); |
188 | ||
189 | if (copy_to_user(arg, &flags, sizeof(flags))) | |
190 | return -EFAULT; | |
191 | return 0; | |
192 | } | |
193 | ||
75e7cb7f LB |
194 | static int check_flags(unsigned int flags) |
195 | { | |
196 | if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ | |
197 | FS_NOATIME_FL | FS_NODUMP_FL | \ | |
198 | FS_SYNC_FL | FS_DIRSYNC_FL | \ | |
e1e8fb6a LZ |
199 | FS_NOCOMP_FL | FS_COMPR_FL | |
200 | FS_NOCOW_FL)) | |
75e7cb7f LB |
201 | return -EOPNOTSUPP; |
202 | ||
203 | if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) | |
204 | return -EINVAL; | |
205 | ||
75e7cb7f LB |
206 | return 0; |
207 | } | |
208 | ||
6cbff00f CH |
209 | static int btrfs_ioctl_setflags(struct file *file, void __user *arg) |
210 | { | |
496ad9aa | 211 | struct inode *inode = file_inode(file); |
6cbff00f CH |
212 | struct btrfs_inode *ip = BTRFS_I(inode); |
213 | struct btrfs_root *root = ip->root; | |
214 | struct btrfs_trans_handle *trans; | |
215 | unsigned int flags, oldflags; | |
216 | int ret; | |
f062abf0 LZ |
217 | u64 ip_oldflags; |
218 | unsigned int i_oldflags; | |
7e97b8da | 219 | umode_t mode; |
6cbff00f | 220 | |
bd60ea0f DS |
221 | if (!inode_owner_or_capable(inode)) |
222 | return -EPERM; | |
223 | ||
b83cc969 LZ |
224 | if (btrfs_root_readonly(root)) |
225 | return -EROFS; | |
226 | ||
6cbff00f CH |
227 | if (copy_from_user(&flags, arg, sizeof(flags))) |
228 | return -EFAULT; | |
229 | ||
75e7cb7f LB |
230 | ret = check_flags(flags); |
231 | if (ret) | |
232 | return ret; | |
f46b5a66 | 233 | |
e7848683 JK |
234 | ret = mnt_want_write_file(file); |
235 | if (ret) | |
236 | return ret; | |
237 | ||
6cbff00f CH |
238 | mutex_lock(&inode->i_mutex); |
239 | ||
f062abf0 LZ |
240 | ip_oldflags = ip->flags; |
241 | i_oldflags = inode->i_flags; | |
7e97b8da | 242 | mode = inode->i_mode; |
f062abf0 | 243 | |
6cbff00f CH |
244 | flags = btrfs_mask_flags(inode->i_mode, flags); |
245 | oldflags = btrfs_flags_to_ioctl(ip->flags); | |
246 | if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { | |
247 | if (!capable(CAP_LINUX_IMMUTABLE)) { | |
248 | ret = -EPERM; | |
249 | goto out_unlock; | |
250 | } | |
251 | } | |
252 | ||
6cbff00f CH |
253 | if (flags & FS_SYNC_FL) |
254 | ip->flags |= BTRFS_INODE_SYNC; | |
255 | else | |
256 | ip->flags &= ~BTRFS_INODE_SYNC; | |
257 | if (flags & FS_IMMUTABLE_FL) | |
258 | ip->flags |= BTRFS_INODE_IMMUTABLE; | |
259 | else | |
260 | ip->flags &= ~BTRFS_INODE_IMMUTABLE; | |
261 | if (flags & FS_APPEND_FL) | |
262 | ip->flags |= BTRFS_INODE_APPEND; | |
263 | else | |
264 | ip->flags &= ~BTRFS_INODE_APPEND; | |
265 | if (flags & FS_NODUMP_FL) | |
266 | ip->flags |= BTRFS_INODE_NODUMP; | |
267 | else | |
268 | ip->flags &= ~BTRFS_INODE_NODUMP; | |
269 | if (flags & FS_NOATIME_FL) | |
270 | ip->flags |= BTRFS_INODE_NOATIME; | |
271 | else | |
272 | ip->flags &= ~BTRFS_INODE_NOATIME; | |
273 | if (flags & FS_DIRSYNC_FL) | |
274 | ip->flags |= BTRFS_INODE_DIRSYNC; | |
275 | else | |
276 | ip->flags &= ~BTRFS_INODE_DIRSYNC; | |
7e97b8da DS |
277 | if (flags & FS_NOCOW_FL) { |
278 | if (S_ISREG(mode)) { | |
279 | /* | |
280 | * It's safe to turn csums off here, no extents exist. | |
281 | * Otherwise we want the flag to reflect the real COW | |
282 | * status of the file and will not set it. | |
283 | */ | |
284 | if (inode->i_size == 0) | |
285 | ip->flags |= BTRFS_INODE_NODATACOW | |
286 | | BTRFS_INODE_NODATASUM; | |
287 | } else { | |
288 | ip->flags |= BTRFS_INODE_NODATACOW; | |
289 | } | |
290 | } else { | |
291 | /* | |
292 | * Revert back under same assuptions as above | |
293 | */ | |
294 | if (S_ISREG(mode)) { | |
295 | if (inode->i_size == 0) | |
296 | ip->flags &= ~(BTRFS_INODE_NODATACOW | |
297 | | BTRFS_INODE_NODATASUM); | |
298 | } else { | |
299 | ip->flags &= ~BTRFS_INODE_NODATACOW; | |
300 | } | |
301 | } | |
6cbff00f | 302 | |
75e7cb7f LB |
303 | /* |
304 | * The COMPRESS flag can only be changed by users, while the NOCOMPRESS | |
305 | * flag may be changed automatically if compression code won't make | |
306 | * things smaller. | |
307 | */ | |
308 | if (flags & FS_NOCOMP_FL) { | |
309 | ip->flags &= ~BTRFS_INODE_COMPRESS; | |
310 | ip->flags |= BTRFS_INODE_NOCOMPRESS; | |
63541927 FDBM |
311 | |
312 | ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); | |
313 | if (ret && ret != -ENODATA) | |
314 | goto out_drop; | |
75e7cb7f | 315 | } else if (flags & FS_COMPR_FL) { |
63541927 FDBM |
316 | const char *comp; |
317 | ||
75e7cb7f LB |
318 | ip->flags |= BTRFS_INODE_COMPRESS; |
319 | ip->flags &= ~BTRFS_INODE_NOCOMPRESS; | |
63541927 FDBM |
320 | |
321 | if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO) | |
322 | comp = "lzo"; | |
323 | else | |
324 | comp = "zlib"; | |
325 | ret = btrfs_set_prop(inode, "btrfs.compression", | |
326 | comp, strlen(comp), 0); | |
327 | if (ret) | |
328 | goto out_drop; | |
329 | ||
ebcb904d LZ |
330 | } else { |
331 | ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); | |
75e7cb7f | 332 | } |
6cbff00f | 333 | |
4da6f1a3 | 334 | trans = btrfs_start_transaction(root, 1); |
f062abf0 LZ |
335 | if (IS_ERR(trans)) { |
336 | ret = PTR_ERR(trans); | |
337 | goto out_drop; | |
338 | } | |
6cbff00f | 339 | |
306424cc | 340 | btrfs_update_iflags(inode); |
0c4d2d95 | 341 | inode_inc_iversion(inode); |
306424cc | 342 | inode->i_ctime = CURRENT_TIME; |
6cbff00f | 343 | ret = btrfs_update_inode(trans, root, inode); |
6cbff00f | 344 | |
6cbff00f | 345 | btrfs_end_transaction(trans, root); |
f062abf0 LZ |
346 | out_drop: |
347 | if (ret) { | |
348 | ip->flags = ip_oldflags; | |
349 | inode->i_flags = i_oldflags; | |
350 | } | |
6cbff00f | 351 | |
6cbff00f CH |
352 | out_unlock: |
353 | mutex_unlock(&inode->i_mutex); | |
e7848683 | 354 | mnt_drop_write_file(file); |
2d4e6f6a | 355 | return ret; |
6cbff00f CH |
356 | } |
357 | ||
358 | static int btrfs_ioctl_getversion(struct file *file, int __user *arg) | |
359 | { | |
496ad9aa | 360 | struct inode *inode = file_inode(file); |
6cbff00f CH |
361 | |
362 | return put_user(inode->i_generation, arg); | |
363 | } | |
f46b5a66 | 364 | |
f7039b1d LD |
365 | static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) |
366 | { | |
54563d41 | 367 | struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); |
f7039b1d LD |
368 | struct btrfs_device *device; |
369 | struct request_queue *q; | |
370 | struct fstrim_range range; | |
371 | u64 minlen = ULLONG_MAX; | |
372 | u64 num_devices = 0; | |
815745cf | 373 | u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy); |
f7039b1d LD |
374 | int ret; |
375 | ||
376 | if (!capable(CAP_SYS_ADMIN)) | |
377 | return -EPERM; | |
378 | ||
1f78160c XG |
379 | rcu_read_lock(); |
380 | list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, | |
381 | dev_list) { | |
f7039b1d LD |
382 | if (!device->bdev) |
383 | continue; | |
384 | q = bdev_get_queue(device->bdev); | |
385 | if (blk_queue_discard(q)) { | |
386 | num_devices++; | |
387 | minlen = min((u64)q->limits.discard_granularity, | |
388 | minlen); | |
389 | } | |
390 | } | |
1f78160c | 391 | rcu_read_unlock(); |
f4c697e6 | 392 | |
f7039b1d LD |
393 | if (!num_devices) |
394 | return -EOPNOTSUPP; | |
f7039b1d LD |
395 | if (copy_from_user(&range, arg, sizeof(range))) |
396 | return -EFAULT; | |
e515c18b LC |
397 | if (range.start > total_bytes || |
398 | range.len < fs_info->sb->s_blocksize) | |
f4c697e6 | 399 | return -EINVAL; |
f7039b1d | 400 | |
f4c697e6 | 401 | range.len = min(range.len, total_bytes - range.start); |
f7039b1d | 402 | range.minlen = max(range.minlen, minlen); |
815745cf | 403 | ret = btrfs_trim_fs(fs_info->tree_root, &range); |
f7039b1d LD |
404 | if (ret < 0) |
405 | return ret; | |
406 | ||
407 | if (copy_to_user(arg, &range, sizeof(range))) | |
408 | return -EFAULT; | |
409 | ||
410 | return 0; | |
411 | } | |
412 | ||
dd5f9615 SB |
413 | int btrfs_is_empty_uuid(u8 *uuid) |
414 | { | |
46e0f66a CM |
415 | int i; |
416 | ||
417 | for (i = 0; i < BTRFS_UUID_SIZE; i++) { | |
418 | if (uuid[i]) | |
419 | return 0; | |
420 | } | |
421 | return 1; | |
dd5f9615 SB |
422 | } |
423 | ||
d5c12070 | 424 | static noinline int create_subvol(struct inode *dir, |
cb8e7090 | 425 | struct dentry *dentry, |
72fd032e | 426 | char *name, int namelen, |
6f72c7e2 | 427 | u64 *async_transid, |
8696c533 | 428 | struct btrfs_qgroup_inherit *inherit) |
f46b5a66 CH |
429 | { |
430 | struct btrfs_trans_handle *trans; | |
431 | struct btrfs_key key; | |
432 | struct btrfs_root_item root_item; | |
433 | struct btrfs_inode_item *inode_item; | |
434 | struct extent_buffer *leaf; | |
d5c12070 | 435 | struct btrfs_root *root = BTRFS_I(dir)->root; |
76dda93c | 436 | struct btrfs_root *new_root; |
d5c12070 | 437 | struct btrfs_block_rsv block_rsv; |
8ea05e3a | 438 | struct timespec cur_time = CURRENT_TIME; |
5662344b | 439 | struct inode *inode; |
f46b5a66 CH |
440 | int ret; |
441 | int err; | |
442 | u64 objectid; | |
443 | u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; | |
3de4586c | 444 | u64 index = 0; |
d5c12070 | 445 | u64 qgroup_reserved; |
8ea05e3a | 446 | uuid_le new_uuid; |
f46b5a66 | 447 | |
581bb050 | 448 | ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid); |
2fbe8c8a | 449 | if (ret) |
a22285a6 | 450 | return ret; |
6a912213 | 451 | |
d5c12070 | 452 | btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); |
9ed74f2d | 453 | /* |
d5c12070 MX |
454 | * The same as the snapshot creation, please see the comment |
455 | * of create_snapshot(). | |
9ed74f2d | 456 | */ |
d5c12070 | 457 | ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, |
dd5f9615 | 458 | 8, &qgroup_reserved, false); |
d5c12070 MX |
459 | if (ret) |
460 | return ret; | |
461 | ||
462 | trans = btrfs_start_transaction(root, 0); | |
463 | if (IS_ERR(trans)) { | |
464 | ret = PTR_ERR(trans); | |
de6e8200 LB |
465 | btrfs_subvolume_release_metadata(root, &block_rsv, |
466 | qgroup_reserved); | |
467 | return ret; | |
d5c12070 MX |
468 | } |
469 | trans->block_rsv = &block_rsv; | |
470 | trans->bytes_reserved = block_rsv.size; | |
f46b5a66 | 471 | |
8696c533 | 472 | ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit); |
6f72c7e2 AJ |
473 | if (ret) |
474 | goto fail; | |
475 | ||
5d4f98a2 | 476 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, |
5581a51a | 477 | 0, objectid, NULL, 0, 0, 0); |
8e8a1e31 JB |
478 | if (IS_ERR(leaf)) { |
479 | ret = PTR_ERR(leaf); | |
480 | goto fail; | |
481 | } | |
f46b5a66 | 482 | |
5d4f98a2 | 483 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
f46b5a66 CH |
484 | btrfs_set_header_bytenr(leaf, leaf->start); |
485 | btrfs_set_header_generation(leaf, trans->transid); | |
5d4f98a2 | 486 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); |
f46b5a66 CH |
487 | btrfs_set_header_owner(leaf, objectid); |
488 | ||
0a4e5586 | 489 | write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(), |
f46b5a66 | 490 | BTRFS_FSID_SIZE); |
5d4f98a2 | 491 | write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, |
b308bc2f | 492 | btrfs_header_chunk_tree_uuid(leaf), |
5d4f98a2 | 493 | BTRFS_UUID_SIZE); |
f46b5a66 CH |
494 | btrfs_mark_buffer_dirty(leaf); |
495 | ||
8ea05e3a AB |
496 | memset(&root_item, 0, sizeof(root_item)); |
497 | ||
f46b5a66 | 498 | inode_item = &root_item.inode; |
3cae210f QW |
499 | btrfs_set_stack_inode_generation(inode_item, 1); |
500 | btrfs_set_stack_inode_size(inode_item, 3); | |
501 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
502 | btrfs_set_stack_inode_nbytes(inode_item, root->leafsize); | |
503 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); | |
f46b5a66 | 504 | |
3cae210f QW |
505 | btrfs_set_root_flags(&root_item, 0); |
506 | btrfs_set_root_limit(&root_item, 0); | |
507 | btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); | |
08fe4db1 | 508 | |
f46b5a66 | 509 | btrfs_set_root_bytenr(&root_item, leaf->start); |
84234f3a | 510 | btrfs_set_root_generation(&root_item, trans->transid); |
f46b5a66 CH |
511 | btrfs_set_root_level(&root_item, 0); |
512 | btrfs_set_root_refs(&root_item, 1); | |
86b9f2ec | 513 | btrfs_set_root_used(&root_item, leaf->len); |
80ff3856 | 514 | btrfs_set_root_last_snapshot(&root_item, 0); |
f46b5a66 | 515 | |
8ea05e3a AB |
516 | btrfs_set_root_generation_v2(&root_item, |
517 | btrfs_root_generation(&root_item)); | |
518 | uuid_le_gen(&new_uuid); | |
519 | memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE); | |
3cae210f QW |
520 | btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec); |
521 | btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec); | |
8ea05e3a AB |
522 | root_item.ctime = root_item.otime; |
523 | btrfs_set_root_ctransid(&root_item, trans->transid); | |
524 | btrfs_set_root_otransid(&root_item, trans->transid); | |
f46b5a66 | 525 | |
925baedd | 526 | btrfs_tree_unlock(leaf); |
f46b5a66 CH |
527 | free_extent_buffer(leaf); |
528 | leaf = NULL; | |
529 | ||
530 | btrfs_set_root_dirid(&root_item, new_dirid); | |
531 | ||
532 | key.objectid = objectid; | |
5d4f98a2 | 533 | key.offset = 0; |
f46b5a66 CH |
534 | btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); |
535 | ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, | |
536 | &root_item); | |
537 | if (ret) | |
538 | goto fail; | |
539 | ||
76dda93c YZ |
540 | key.offset = (u64)-1; |
541 | new_root = btrfs_read_fs_root_no_name(root->fs_info, &key); | |
79787eaa JM |
542 | if (IS_ERR(new_root)) { |
543 | btrfs_abort_transaction(trans, root, PTR_ERR(new_root)); | |
544 | ret = PTR_ERR(new_root); | |
545 | goto fail; | |
546 | } | |
76dda93c YZ |
547 | |
548 | btrfs_record_root_in_trans(trans, new_root); | |
549 | ||
63541927 | 550 | ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); |
ce598979 MF |
551 | if (ret) { |
552 | /* We potentially lose an unused inode item here */ | |
79787eaa | 553 | btrfs_abort_transaction(trans, root, ret); |
ce598979 MF |
554 | goto fail; |
555 | } | |
556 | ||
f46b5a66 CH |
557 | /* |
558 | * insert the directory item | |
559 | */ | |
3de4586c | 560 | ret = btrfs_set_inode_index(dir, &index); |
79787eaa JM |
561 | if (ret) { |
562 | btrfs_abort_transaction(trans, root, ret); | |
563 | goto fail; | |
564 | } | |
3de4586c CM |
565 | |
566 | ret = btrfs_insert_dir_item(trans, root, | |
16cdcec7 | 567 | name, namelen, dir, &key, |
3de4586c | 568 | BTRFS_FT_DIR, index); |
79787eaa JM |
569 | if (ret) { |
570 | btrfs_abort_transaction(trans, root, ret); | |
f46b5a66 | 571 | goto fail; |
79787eaa | 572 | } |
0660b5af | 573 | |
52c26179 YZ |
574 | btrfs_i_size_write(dir, dir->i_size + namelen * 2); |
575 | ret = btrfs_update_inode(trans, root, dir); | |
576 | BUG_ON(ret); | |
577 | ||
0660b5af | 578 | ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, |
4df27c4d | 579 | objectid, root->root_key.objectid, |
33345d01 | 580 | btrfs_ino(dir), index, name, namelen); |
76dda93c | 581 | BUG_ON(ret); |
f46b5a66 | 582 | |
dd5f9615 SB |
583 | ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root, |
584 | root_item.uuid, BTRFS_UUID_KEY_SUBVOL, | |
585 | objectid); | |
586 | if (ret) | |
587 | btrfs_abort_transaction(trans, root, ret); | |
588 | ||
f46b5a66 | 589 | fail: |
d5c12070 MX |
590 | trans->block_rsv = NULL; |
591 | trans->bytes_reserved = 0; | |
de6e8200 LB |
592 | btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); |
593 | ||
72fd032e SW |
594 | if (async_transid) { |
595 | *async_transid = trans->transid; | |
596 | err = btrfs_commit_transaction_async(trans, root, 1); | |
00d71c9c MX |
597 | if (err) |
598 | err = btrfs_commit_transaction(trans, root); | |
72fd032e SW |
599 | } else { |
600 | err = btrfs_commit_transaction(trans, root); | |
601 | } | |
f46b5a66 CH |
602 | if (err && !ret) |
603 | ret = err; | |
1a65e24b | 604 | |
5662344b TI |
605 | if (!ret) { |
606 | inode = btrfs_lookup_dentry(dir, dentry); | |
de6e8200 LB |
607 | if (IS_ERR(inode)) |
608 | return PTR_ERR(inode); | |
5662344b TI |
609 | d_instantiate(dentry, inode); |
610 | } | |
f46b5a66 CH |
611 | return ret; |
612 | } | |
613 | ||
8257b2dc MX |
614 | static void btrfs_wait_nocow_write(struct btrfs_root *root) |
615 | { | |
616 | s64 writers; | |
617 | DEFINE_WAIT(wait); | |
618 | ||
619 | do { | |
620 | prepare_to_wait(&root->subv_writers->wait, &wait, | |
621 | TASK_UNINTERRUPTIBLE); | |
622 | ||
623 | writers = percpu_counter_sum(&root->subv_writers->counter); | |
624 | if (writers) | |
625 | schedule(); | |
626 | ||
627 | finish_wait(&root->subv_writers->wait, &wait); | |
628 | } while (writers); | |
629 | } | |
630 | ||
e9662f70 MX |
631 | static int create_snapshot(struct btrfs_root *root, struct inode *dir, |
632 | struct dentry *dentry, char *name, int namelen, | |
633 | u64 *async_transid, bool readonly, | |
634 | struct btrfs_qgroup_inherit *inherit) | |
f46b5a66 | 635 | { |
2e4bfab9 | 636 | struct inode *inode; |
f46b5a66 CH |
637 | struct btrfs_pending_snapshot *pending_snapshot; |
638 | struct btrfs_trans_handle *trans; | |
2e4bfab9 | 639 | int ret; |
f46b5a66 CH |
640 | |
641 | if (!root->ref_cows) | |
642 | return -EINVAL; | |
643 | ||
8257b2dc MX |
644 | atomic_inc(&root->will_be_snapshoted); |
645 | smp_mb__after_atomic_inc(); | |
646 | btrfs_wait_nocow_write(root); | |
647 | ||
6a03843d MX |
648 | ret = btrfs_start_delalloc_inodes(root, 0); |
649 | if (ret) | |
8257b2dc | 650 | goto out; |
6a03843d | 651 | |
b0244199 | 652 | btrfs_wait_ordered_extents(root, -1); |
6a03843d | 653 | |
3de4586c | 654 | pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); |
8257b2dc MX |
655 | if (!pending_snapshot) { |
656 | ret = -ENOMEM; | |
657 | goto out; | |
658 | } | |
a22285a6 | 659 | |
66d8f3dd MX |
660 | btrfs_init_block_rsv(&pending_snapshot->block_rsv, |
661 | BTRFS_BLOCK_RSV_TEMP); | |
d5c12070 MX |
662 | /* |
663 | * 1 - parent dir inode | |
664 | * 2 - dir entries | |
665 | * 1 - root item | |
666 | * 2 - root ref/backref | |
667 | * 1 - root of snapshot | |
dd5f9615 | 668 | * 1 - UUID item |
d5c12070 MX |
669 | */ |
670 | ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, | |
dd5f9615 | 671 | &pending_snapshot->block_rsv, 8, |
ee3441b4 JM |
672 | &pending_snapshot->qgroup_reserved, |
673 | false); | |
d5c12070 | 674 | if (ret) |
8257b2dc | 675 | goto free; |
d5c12070 | 676 | |
3de4586c | 677 | pending_snapshot->dentry = dentry; |
f46b5a66 | 678 | pending_snapshot->root = root; |
b83cc969 | 679 | pending_snapshot->readonly = readonly; |
e9662f70 | 680 | pending_snapshot->dir = dir; |
8696c533 | 681 | pending_snapshot->inherit = inherit; |
a22285a6 | 682 | |
d5c12070 | 683 | trans = btrfs_start_transaction(root, 0); |
a22285a6 YZ |
684 | if (IS_ERR(trans)) { |
685 | ret = PTR_ERR(trans); | |
686 | goto fail; | |
687 | } | |
688 | ||
8351583e | 689 | spin_lock(&root->fs_info->trans_lock); |
f46b5a66 CH |
690 | list_add(&pending_snapshot->list, |
691 | &trans->transaction->pending_snapshots); | |
8351583e | 692 | spin_unlock(&root->fs_info->trans_lock); |
72fd032e SW |
693 | if (async_transid) { |
694 | *async_transid = trans->transid; | |
695 | ret = btrfs_commit_transaction_async(trans, | |
696 | root->fs_info->extent_root, 1); | |
00d71c9c MX |
697 | if (ret) |
698 | ret = btrfs_commit_transaction(trans, root); | |
72fd032e SW |
699 | } else { |
700 | ret = btrfs_commit_transaction(trans, | |
701 | root->fs_info->extent_root); | |
702 | } | |
aec8030a | 703 | if (ret) |
c37b2b62 | 704 | goto fail; |
a22285a6 YZ |
705 | |
706 | ret = pending_snapshot->error; | |
707 | if (ret) | |
708 | goto fail; | |
709 | ||
66b4ffd1 JB |
710 | ret = btrfs_orphan_cleanup(pending_snapshot->snap); |
711 | if (ret) | |
712 | goto fail; | |
f46b5a66 | 713 | |
2fbe8c8a | 714 | inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry); |
2e4bfab9 YZ |
715 | if (IS_ERR(inode)) { |
716 | ret = PTR_ERR(inode); | |
717 | goto fail; | |
718 | } | |
5662344b | 719 | |
2e4bfab9 YZ |
720 | d_instantiate(dentry, inode); |
721 | ret = 0; | |
722 | fail: | |
d5c12070 MX |
723 | btrfs_subvolume_release_metadata(BTRFS_I(dir)->root, |
724 | &pending_snapshot->block_rsv, | |
725 | pending_snapshot->qgroup_reserved); | |
8257b2dc | 726 | free: |
a22285a6 | 727 | kfree(pending_snapshot); |
8257b2dc MX |
728 | out: |
729 | atomic_dec(&root->will_be_snapshoted); | |
f46b5a66 CH |
730 | return ret; |
731 | } | |
732 | ||
4260f7c7 SW |
733 | /* copy of check_sticky in fs/namei.c() |
734 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
735 | * minimal. | |
736 | */ | |
737 | static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode) | |
738 | { | |
2f2f43d3 | 739 | kuid_t fsuid = current_fsuid(); |
4260f7c7 SW |
740 | |
741 | if (!(dir->i_mode & S_ISVTX)) | |
742 | return 0; | |
2f2f43d3 | 743 | if (uid_eq(inode->i_uid, fsuid)) |
4260f7c7 | 744 | return 0; |
2f2f43d3 | 745 | if (uid_eq(dir->i_uid, fsuid)) |
4260f7c7 SW |
746 | return 0; |
747 | return !capable(CAP_FOWNER); | |
748 | } | |
749 | ||
750 | /* copy of may_delete in fs/namei.c() | |
751 | * Check whether we can remove a link victim from directory dir, check | |
752 | * whether the type of victim is right. | |
753 | * 1. We can't do it if dir is read-only (done in permission()) | |
754 | * 2. We should have write and exec permissions on dir | |
755 | * 3. We can't remove anything from append-only dir | |
756 | * 4. We can't do anything with immutable dir (done in permission()) | |
757 | * 5. If the sticky bit on dir is set we should either | |
758 | * a. be owner of dir, or | |
759 | * b. be owner of victim, or | |
760 | * c. have CAP_FOWNER capability | |
761 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
762 | * links pointing to it. | |
763 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
764 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
765 | * 9. We can't remove a root or mountpoint. | |
766 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
767 | * nfs_async_unlink(). | |
768 | */ | |
769 | ||
67871254 | 770 | static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) |
4260f7c7 SW |
771 | { |
772 | int error; | |
773 | ||
774 | if (!victim->d_inode) | |
775 | return -ENOENT; | |
776 | ||
777 | BUG_ON(victim->d_parent->d_inode != dir); | |
4fa6b5ec | 778 | audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); |
4260f7c7 SW |
779 | |
780 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); | |
781 | if (error) | |
782 | return error; | |
783 | if (IS_APPEND(dir)) | |
784 | return -EPERM; | |
785 | if (btrfs_check_sticky(dir, victim->d_inode)|| | |
786 | IS_APPEND(victim->d_inode)|| | |
787 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) | |
788 | return -EPERM; | |
789 | if (isdir) { | |
790 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
791 | return -ENOTDIR; | |
792 | if (IS_ROOT(victim)) | |
793 | return -EBUSY; | |
794 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
795 | return -EISDIR; | |
796 | if (IS_DEADDIR(dir)) | |
797 | return -ENOENT; | |
798 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
799 | return -EBUSY; | |
800 | return 0; | |
801 | } | |
802 | ||
cb8e7090 CH |
803 | /* copy of may_create in fs/namei.c() */ |
804 | static inline int btrfs_may_create(struct inode *dir, struct dentry *child) | |
805 | { | |
806 | if (child->d_inode) | |
807 | return -EEXIST; | |
808 | if (IS_DEADDIR(dir)) | |
809 | return -ENOENT; | |
810 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); | |
811 | } | |
812 | ||
813 | /* | |
814 | * Create a new subvolume below @parent. This is largely modeled after | |
815 | * sys_mkdirat and vfs_mkdir, but we only do a single component lookup | |
816 | * inside this filesystem so it's quite a bit simpler. | |
817 | */ | |
76dda93c YZ |
818 | static noinline int btrfs_mksubvol(struct path *parent, |
819 | char *name, int namelen, | |
72fd032e | 820 | struct btrfs_root *snap_src, |
6f72c7e2 | 821 | u64 *async_transid, bool readonly, |
8696c533 | 822 | struct btrfs_qgroup_inherit *inherit) |
cb8e7090 | 823 | { |
76dda93c | 824 | struct inode *dir = parent->dentry->d_inode; |
cb8e7090 CH |
825 | struct dentry *dentry; |
826 | int error; | |
827 | ||
5c50c9b8 DS |
828 | error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT); |
829 | if (error == -EINTR) | |
830 | return error; | |
cb8e7090 CH |
831 | |
832 | dentry = lookup_one_len(name, parent->dentry, namelen); | |
833 | error = PTR_ERR(dentry); | |
834 | if (IS_ERR(dentry)) | |
835 | goto out_unlock; | |
836 | ||
837 | error = -EEXIST; | |
838 | if (dentry->d_inode) | |
839 | goto out_dput; | |
840 | ||
76dda93c | 841 | error = btrfs_may_create(dir, dentry); |
cb8e7090 | 842 | if (error) |
a874a63e | 843 | goto out_dput; |
cb8e7090 | 844 | |
9c52057c CM |
845 | /* |
846 | * even if this name doesn't exist, we may get hash collisions. | |
847 | * check for them now when we can safely fail | |
848 | */ | |
849 | error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, | |
850 | dir->i_ino, name, | |
851 | namelen); | |
852 | if (error) | |
853 | goto out_dput; | |
854 | ||
76dda93c YZ |
855 | down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); |
856 | ||
857 | if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) | |
858 | goto out_up_read; | |
859 | ||
3de4586c | 860 | if (snap_src) { |
e9662f70 | 861 | error = create_snapshot(snap_src, dir, dentry, name, namelen, |
6f72c7e2 | 862 | async_transid, readonly, inherit); |
3de4586c | 863 | } else { |
d5c12070 MX |
864 | error = create_subvol(dir, dentry, name, namelen, |
865 | async_transid, inherit); | |
3de4586c | 866 | } |
76dda93c YZ |
867 | if (!error) |
868 | fsnotify_mkdir(dir, dentry); | |
869 | out_up_read: | |
870 | up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); | |
cb8e7090 CH |
871 | out_dput: |
872 | dput(dentry); | |
873 | out_unlock: | |
76dda93c | 874 | mutex_unlock(&dir->i_mutex); |
cb8e7090 CH |
875 | return error; |
876 | } | |
877 | ||
4cb5300b CM |
878 | /* |
879 | * When we're defragging a range, we don't want to kick it off again | |
880 | * if it is really just waiting for delalloc to send it down. | |
881 | * If we find a nice big extent or delalloc range for the bytes in the | |
882 | * file you want to defrag, we return 0 to let you know to skip this | |
883 | * part of the file | |
884 | */ | |
885 | static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh) | |
886 | { | |
887 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | |
888 | struct extent_map *em = NULL; | |
889 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
890 | u64 end; | |
891 | ||
892 | read_lock(&em_tree->lock); | |
893 | em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); | |
894 | read_unlock(&em_tree->lock); | |
895 | ||
896 | if (em) { | |
897 | end = extent_map_end(em); | |
898 | free_extent_map(em); | |
899 | if (end - offset > thresh) | |
900 | return 0; | |
901 | } | |
902 | /* if we already have a nice delalloc here, just stop */ | |
903 | thresh /= 2; | |
904 | end = count_range_bits(io_tree, &offset, offset + thresh, | |
905 | thresh, EXTENT_DELALLOC, 1); | |
906 | if (end >= thresh) | |
907 | return 0; | |
908 | return 1; | |
909 | } | |
910 | ||
911 | /* | |
912 | * helper function to walk through a file and find extents | |
913 | * newer than a specific transid, and smaller than thresh. | |
914 | * | |
915 | * This is used by the defragging code to find new and small | |
916 | * extents | |
917 | */ | |
918 | static int find_new_extents(struct btrfs_root *root, | |
919 | struct inode *inode, u64 newer_than, | |
920 | u64 *off, int thresh) | |
921 | { | |
922 | struct btrfs_path *path; | |
923 | struct btrfs_key min_key; | |
4cb5300b CM |
924 | struct extent_buffer *leaf; |
925 | struct btrfs_file_extent_item *extent; | |
926 | int type; | |
927 | int ret; | |
a4689d2b | 928 | u64 ino = btrfs_ino(inode); |
4cb5300b CM |
929 | |
930 | path = btrfs_alloc_path(); | |
931 | if (!path) | |
932 | return -ENOMEM; | |
933 | ||
a4689d2b | 934 | min_key.objectid = ino; |
4cb5300b CM |
935 | min_key.type = BTRFS_EXTENT_DATA_KEY; |
936 | min_key.offset = *off; | |
937 | ||
67871254 | 938 | while (1) { |
f094c9bd | 939 | path->keep_locks = 1; |
6174d3cb | 940 | ret = btrfs_search_forward(root, &min_key, path, newer_than); |
4cb5300b CM |
941 | if (ret != 0) |
942 | goto none; | |
f094c9bd FM |
943 | path->keep_locks = 0; |
944 | btrfs_unlock_up_safe(path, 1); | |
945 | process_slot: | |
a4689d2b | 946 | if (min_key.objectid != ino) |
4cb5300b CM |
947 | goto none; |
948 | if (min_key.type != BTRFS_EXTENT_DATA_KEY) | |
949 | goto none; | |
950 | ||
951 | leaf = path->nodes[0]; | |
952 | extent = btrfs_item_ptr(leaf, path->slots[0], | |
953 | struct btrfs_file_extent_item); | |
954 | ||
955 | type = btrfs_file_extent_type(leaf, extent); | |
956 | if (type == BTRFS_FILE_EXTENT_REG && | |
957 | btrfs_file_extent_num_bytes(leaf, extent) < thresh && | |
958 | check_defrag_in_cache(inode, min_key.offset, thresh)) { | |
959 | *off = min_key.offset; | |
960 | btrfs_free_path(path); | |
961 | return 0; | |
962 | } | |
963 | ||
f094c9bd FM |
964 | path->slots[0]++; |
965 | if (path->slots[0] < btrfs_header_nritems(leaf)) { | |
966 | btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); | |
967 | goto process_slot; | |
968 | } | |
969 | ||
4cb5300b CM |
970 | if (min_key.offset == (u64)-1) |
971 | goto none; | |
972 | ||
973 | min_key.offset++; | |
974 | btrfs_release_path(path); | |
975 | } | |
976 | none: | |
977 | btrfs_free_path(path); | |
978 | return -ENOENT; | |
979 | } | |
980 | ||
6c282eb4 | 981 | static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) |
17ce6ef8 LB |
982 | { |
983 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
6c282eb4 LZ |
984 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
985 | struct extent_map *em; | |
986 | u64 len = PAGE_CACHE_SIZE; | |
17ce6ef8 | 987 | |
6c282eb4 LZ |
988 | /* |
989 | * hopefully we have this extent in the tree already, try without | |
990 | * the full extent lock | |
991 | */ | |
17ce6ef8 | 992 | read_lock(&em_tree->lock); |
6c282eb4 | 993 | em = lookup_extent_mapping(em_tree, start, len); |
17ce6ef8 LB |
994 | read_unlock(&em_tree->lock); |
995 | ||
6c282eb4 | 996 | if (!em) { |
308d9800 FM |
997 | struct extent_state *cached = NULL; |
998 | u64 end = start + len - 1; | |
999 | ||
6c282eb4 | 1000 | /* get the big lock and read metadata off disk */ |
308d9800 | 1001 | lock_extent_bits(io_tree, start, end, 0, &cached); |
6c282eb4 | 1002 | em = btrfs_get_extent(inode, NULL, 0, start, len, 0); |
308d9800 | 1003 | unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS); |
6c282eb4 LZ |
1004 | |
1005 | if (IS_ERR(em)) | |
1006 | return NULL; | |
1007 | } | |
1008 | ||
1009 | return em; | |
1010 | } | |
17ce6ef8 | 1011 | |
6c282eb4 LZ |
1012 | static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) |
1013 | { | |
1014 | struct extent_map *next; | |
1015 | bool ret = true; | |
1016 | ||
1017 | /* this is the last extent */ | |
1018 | if (em->start + em->len >= i_size_read(inode)) | |
1019 | return false; | |
1020 | ||
1021 | next = defrag_lookup_extent(inode, em->start + em->len); | |
e2127cf0 FM |
1022 | if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE || |
1023 | (em->block_start + em->block_len == next->block_start)) | |
6c282eb4 LZ |
1024 | ret = false; |
1025 | ||
1026 | free_extent_map(next); | |
17ce6ef8 LB |
1027 | return ret; |
1028 | } | |
1029 | ||
6c282eb4 | 1030 | static int should_defrag_range(struct inode *inode, u64 start, int thresh, |
a43a2111 AM |
1031 | u64 *last_len, u64 *skip, u64 *defrag_end, |
1032 | int compress) | |
940100a4 | 1033 | { |
6c282eb4 | 1034 | struct extent_map *em; |
940100a4 | 1035 | int ret = 1; |
6c282eb4 | 1036 | bool next_mergeable = true; |
940100a4 CM |
1037 | |
1038 | /* | |
008873ea | 1039 | * make sure that once we start defragging an extent, we keep on |
940100a4 CM |
1040 | * defragging it |
1041 | */ | |
1042 | if (start < *defrag_end) | |
1043 | return 1; | |
1044 | ||
1045 | *skip = 0; | |
1046 | ||
6c282eb4 LZ |
1047 | em = defrag_lookup_extent(inode, start); |
1048 | if (!em) | |
1049 | return 0; | |
940100a4 CM |
1050 | |
1051 | /* this will cover holes, and inline extents */ | |
17ce6ef8 | 1052 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) { |
940100a4 | 1053 | ret = 0; |
17ce6ef8 LB |
1054 | goto out; |
1055 | } | |
1056 | ||
6c282eb4 | 1057 | next_mergeable = defrag_check_next_extent(inode, em); |
940100a4 CM |
1058 | |
1059 | /* | |
6c282eb4 LZ |
1060 | * we hit a real extent, if it is big or the next extent is not a |
1061 | * real extent, don't bother defragging it | |
940100a4 | 1062 | */ |
a43a2111 | 1063 | if (!compress && (*last_len == 0 || *last_len >= thresh) && |
6c282eb4 | 1064 | (em->len >= thresh || !next_mergeable)) |
940100a4 | 1065 | ret = 0; |
17ce6ef8 | 1066 | out: |
940100a4 CM |
1067 | /* |
1068 | * last_len ends up being a counter of how many bytes we've defragged. | |
1069 | * every time we choose not to defrag an extent, we reset *last_len | |
1070 | * so that the next tiny extent will force a defrag. | |
1071 | * | |
1072 | * The end result of this is that tiny extents before a single big | |
1073 | * extent will force at least part of that big extent to be defragged. | |
1074 | */ | |
1075 | if (ret) { | |
940100a4 CM |
1076 | *defrag_end = extent_map_end(em); |
1077 | } else { | |
1078 | *last_len = 0; | |
1079 | *skip = extent_map_end(em); | |
1080 | *defrag_end = 0; | |
1081 | } | |
1082 | ||
1083 | free_extent_map(em); | |
1084 | return ret; | |
1085 | } | |
1086 | ||
4cb5300b CM |
1087 | /* |
1088 | * it doesn't do much good to defrag one or two pages | |
1089 | * at a time. This pulls in a nice chunk of pages | |
1090 | * to COW and defrag. | |
1091 | * | |
1092 | * It also makes sure the delalloc code has enough | |
1093 | * dirty data to avoid making new small extents as part | |
1094 | * of the defrag | |
1095 | * | |
1096 | * It's a good idea to start RA on this range | |
1097 | * before calling this. | |
1098 | */ | |
1099 | static int cluster_pages_for_defrag(struct inode *inode, | |
1100 | struct page **pages, | |
1101 | unsigned long start_index, | |
c41570c9 | 1102 | unsigned long num_pages) |
f46b5a66 | 1103 | { |
4cb5300b CM |
1104 | unsigned long file_end; |
1105 | u64 isize = i_size_read(inode); | |
1106 | u64 page_start; | |
1107 | u64 page_end; | |
1f12bd06 | 1108 | u64 page_cnt; |
4cb5300b CM |
1109 | int ret; |
1110 | int i; | |
1111 | int i_done; | |
3eaa2885 | 1112 | struct btrfs_ordered_extent *ordered; |
4cb5300b | 1113 | struct extent_state *cached_state = NULL; |
600a45e1 | 1114 | struct extent_io_tree *tree; |
3b16a4e3 | 1115 | gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); |
4cb5300b | 1116 | |
4cb5300b | 1117 | file_end = (isize - 1) >> PAGE_CACHE_SHIFT; |
1f12bd06 LB |
1118 | if (!isize || start_index > file_end) |
1119 | return 0; | |
1120 | ||
1121 | page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); | |
4cb5300b CM |
1122 | |
1123 | ret = btrfs_delalloc_reserve_space(inode, | |
1f12bd06 | 1124 | page_cnt << PAGE_CACHE_SHIFT); |
4cb5300b CM |
1125 | if (ret) |
1126 | return ret; | |
4cb5300b | 1127 | i_done = 0; |
600a45e1 | 1128 | tree = &BTRFS_I(inode)->io_tree; |
4cb5300b CM |
1129 | |
1130 | /* step one, lock all the pages */ | |
1f12bd06 | 1131 | for (i = 0; i < page_cnt; i++) { |
4cb5300b | 1132 | struct page *page; |
600a45e1 | 1133 | again: |
a94733d0 | 1134 | page = find_or_create_page(inode->i_mapping, |
600a45e1 | 1135 | start_index + i, mask); |
4cb5300b CM |
1136 | if (!page) |
1137 | break; | |
1138 | ||
600a45e1 MX |
1139 | page_start = page_offset(page); |
1140 | page_end = page_start + PAGE_CACHE_SIZE - 1; | |
1141 | while (1) { | |
308d9800 FM |
1142 | lock_extent_bits(tree, page_start, page_end, |
1143 | 0, &cached_state); | |
600a45e1 MX |
1144 | ordered = btrfs_lookup_ordered_extent(inode, |
1145 | page_start); | |
308d9800 FM |
1146 | unlock_extent_cached(tree, page_start, page_end, |
1147 | &cached_state, GFP_NOFS); | |
600a45e1 MX |
1148 | if (!ordered) |
1149 | break; | |
1150 | ||
1151 | unlock_page(page); | |
1152 | btrfs_start_ordered_extent(inode, ordered, 1); | |
1153 | btrfs_put_ordered_extent(ordered); | |
1154 | lock_page(page); | |
1f12bd06 LB |
1155 | /* |
1156 | * we unlocked the page above, so we need check if | |
1157 | * it was released or not. | |
1158 | */ | |
1159 | if (page->mapping != inode->i_mapping) { | |
1160 | unlock_page(page); | |
1161 | page_cache_release(page); | |
1162 | goto again; | |
1163 | } | |
600a45e1 MX |
1164 | } |
1165 | ||
4cb5300b CM |
1166 | if (!PageUptodate(page)) { |
1167 | btrfs_readpage(NULL, page); | |
1168 | lock_page(page); | |
1169 | if (!PageUptodate(page)) { | |
1170 | unlock_page(page); | |
1171 | page_cache_release(page); | |
1172 | ret = -EIO; | |
1173 | break; | |
1174 | } | |
1175 | } | |
600a45e1 | 1176 | |
600a45e1 MX |
1177 | if (page->mapping != inode->i_mapping) { |
1178 | unlock_page(page); | |
1179 | page_cache_release(page); | |
1180 | goto again; | |
1181 | } | |
1182 | ||
4cb5300b CM |
1183 | pages[i] = page; |
1184 | i_done++; | |
1185 | } | |
1186 | if (!i_done || ret) | |
1187 | goto out; | |
1188 | ||
1189 | if (!(inode->i_sb->s_flags & MS_ACTIVE)) | |
1190 | goto out; | |
1191 | ||
1192 | /* | |
1193 | * so now we have a nice long stream of locked | |
1194 | * and up to date pages, lets wait on them | |
1195 | */ | |
1196 | for (i = 0; i < i_done; i++) | |
1197 | wait_on_page_writeback(pages[i]); | |
1198 | ||
1199 | page_start = page_offset(pages[0]); | |
1200 | page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE; | |
1201 | ||
1202 | lock_extent_bits(&BTRFS_I(inode)->io_tree, | |
d0082371 | 1203 | page_start, page_end - 1, 0, &cached_state); |
4cb5300b CM |
1204 | clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, |
1205 | page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | | |
9e8a4a8b LB |
1206 | EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, |
1207 | &cached_state, GFP_NOFS); | |
4cb5300b | 1208 | |
1f12bd06 | 1209 | if (i_done != page_cnt) { |
9e0baf60 JB |
1210 | spin_lock(&BTRFS_I(inode)->lock); |
1211 | BTRFS_I(inode)->outstanding_extents++; | |
1212 | spin_unlock(&BTRFS_I(inode)->lock); | |
4cb5300b | 1213 | btrfs_delalloc_release_space(inode, |
1f12bd06 | 1214 | (page_cnt - i_done) << PAGE_CACHE_SHIFT); |
4cb5300b CM |
1215 | } |
1216 | ||
1217 | ||
9e8a4a8b LB |
1218 | set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, |
1219 | &cached_state, GFP_NOFS); | |
4cb5300b CM |
1220 | |
1221 | unlock_extent_cached(&BTRFS_I(inode)->io_tree, | |
1222 | page_start, page_end - 1, &cached_state, | |
1223 | GFP_NOFS); | |
1224 | ||
1225 | for (i = 0; i < i_done; i++) { | |
1226 | clear_page_dirty_for_io(pages[i]); | |
1227 | ClearPageChecked(pages[i]); | |
1228 | set_page_extent_mapped(pages[i]); | |
1229 | set_page_dirty(pages[i]); | |
1230 | unlock_page(pages[i]); | |
1231 | page_cache_release(pages[i]); | |
1232 | } | |
1233 | return i_done; | |
1234 | out: | |
1235 | for (i = 0; i < i_done; i++) { | |
1236 | unlock_page(pages[i]); | |
1237 | page_cache_release(pages[i]); | |
1238 | } | |
1f12bd06 | 1239 | btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT); |
4cb5300b CM |
1240 | return ret; |
1241 | ||
1242 | } | |
1243 | ||
1244 | int btrfs_defrag_file(struct inode *inode, struct file *file, | |
1245 | struct btrfs_ioctl_defrag_range_args *range, | |
1246 | u64 newer_than, unsigned long max_to_defrag) | |
1247 | { | |
1248 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
4cb5300b | 1249 | struct file_ra_state *ra = NULL; |
f46b5a66 | 1250 | unsigned long last_index; |
151a31b2 | 1251 | u64 isize = i_size_read(inode); |
940100a4 CM |
1252 | u64 last_len = 0; |
1253 | u64 skip = 0; | |
1254 | u64 defrag_end = 0; | |
4cb5300b | 1255 | u64 newer_off = range->start; |
f46b5a66 | 1256 | unsigned long i; |
008873ea | 1257 | unsigned long ra_index = 0; |
f46b5a66 | 1258 | int ret; |
4cb5300b | 1259 | int defrag_count = 0; |
1a419d85 | 1260 | int compress_type = BTRFS_COMPRESS_ZLIB; |
4cb5300b | 1261 | int extent_thresh = range->extent_thresh; |
c41570c9 JM |
1262 | unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT; |
1263 | unsigned long cluster = max_cluster; | |
4cb5300b CM |
1264 | u64 new_align = ~((u64)128 * 1024 - 1); |
1265 | struct page **pages = NULL; | |
1266 | ||
0abd5b17 LB |
1267 | if (isize == 0) |
1268 | return 0; | |
1269 | ||
1270 | if (range->start >= isize) | |
1271 | return -EINVAL; | |
1a419d85 LZ |
1272 | |
1273 | if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { | |
1274 | if (range->compress_type > BTRFS_COMPRESS_TYPES) | |
1275 | return -EINVAL; | |
1276 | if (range->compress_type) | |
1277 | compress_type = range->compress_type; | |
1278 | } | |
f46b5a66 | 1279 | |
0abd5b17 LB |
1280 | if (extent_thresh == 0) |
1281 | extent_thresh = 256 * 1024; | |
940100a4 | 1282 | |
4cb5300b CM |
1283 | /* |
1284 | * if we were not given a file, allocate a readahead | |
1285 | * context | |
1286 | */ | |
1287 | if (!file) { | |
1288 | ra = kzalloc(sizeof(*ra), GFP_NOFS); | |
1289 | if (!ra) | |
1290 | return -ENOMEM; | |
1291 | file_ra_state_init(ra, inode->i_mapping); | |
1292 | } else { | |
1293 | ra = &file->f_ra; | |
1294 | } | |
1295 | ||
d9b0d9ba | 1296 | pages = kmalloc_array(max_cluster, sizeof(struct page *), |
4cb5300b CM |
1297 | GFP_NOFS); |
1298 | if (!pages) { | |
1299 | ret = -ENOMEM; | |
1300 | goto out_ra; | |
1301 | } | |
1302 | ||
1303 | /* find the last page to defrag */ | |
1e701a32 | 1304 | if (range->start + range->len > range->start) { |
151a31b2 | 1305 | last_index = min_t(u64, isize - 1, |
1e701a32 CM |
1306 | range->start + range->len - 1) >> PAGE_CACHE_SHIFT; |
1307 | } else { | |
151a31b2 | 1308 | last_index = (isize - 1) >> PAGE_CACHE_SHIFT; |
1e701a32 CM |
1309 | } |
1310 | ||
4cb5300b CM |
1311 | if (newer_than) { |
1312 | ret = find_new_extents(root, inode, newer_than, | |
1313 | &newer_off, 64 * 1024); | |
1314 | if (!ret) { | |
1315 | range->start = newer_off; | |
1316 | /* | |
1317 | * we always align our defrag to help keep | |
1318 | * the extents in the file evenly spaced | |
1319 | */ | |
1320 | i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; | |
4cb5300b CM |
1321 | } else |
1322 | goto out_ra; | |
1323 | } else { | |
1324 | i = range->start >> PAGE_CACHE_SHIFT; | |
1325 | } | |
1326 | if (!max_to_defrag) | |
7ec31b54 | 1327 | max_to_defrag = last_index + 1; |
4cb5300b | 1328 | |
2a0f7f57 LZ |
1329 | /* |
1330 | * make writeback starts from i, so the defrag range can be | |
1331 | * written sequentially. | |
1332 | */ | |
1333 | if (i < inode->i_mapping->writeback_index) | |
1334 | inode->i_mapping->writeback_index = i; | |
1335 | ||
f7f43cc8 CM |
1336 | while (i <= last_index && defrag_count < max_to_defrag && |
1337 | (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >> | |
1338 | PAGE_CACHE_SHIFT)) { | |
4cb5300b CM |
1339 | /* |
1340 | * make sure we stop running if someone unmounts | |
1341 | * the FS | |
1342 | */ | |
1343 | if (!(inode->i_sb->s_flags & MS_ACTIVE)) | |
1344 | break; | |
1345 | ||
210549eb | 1346 | if (btrfs_defrag_cancelled(root->fs_info)) { |
efe120a0 | 1347 | printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n"); |
210549eb DS |
1348 | ret = -EAGAIN; |
1349 | break; | |
1350 | } | |
1351 | ||
66c26892 | 1352 | if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT, |
6c282eb4 | 1353 | extent_thresh, &last_len, &skip, |
a43a2111 AM |
1354 | &defrag_end, range->flags & |
1355 | BTRFS_DEFRAG_RANGE_COMPRESS)) { | |
940100a4 CM |
1356 | unsigned long next; |
1357 | /* | |
1358 | * the should_defrag function tells us how much to skip | |
1359 | * bump our counter by the suggested amount | |
1360 | */ | |
1361 | next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; | |
1362 | i = max(i + 1, next); | |
1363 | continue; | |
1364 | } | |
008873ea LZ |
1365 | |
1366 | if (!newer_than) { | |
1367 | cluster = (PAGE_CACHE_ALIGN(defrag_end) >> | |
1368 | PAGE_CACHE_SHIFT) - i; | |
1369 | cluster = min(cluster, max_cluster); | |
1370 | } else { | |
1371 | cluster = max_cluster; | |
1372 | } | |
1373 | ||
008873ea LZ |
1374 | if (i + cluster > ra_index) { |
1375 | ra_index = max(i, ra_index); | |
1376 | btrfs_force_ra(inode->i_mapping, ra, file, ra_index, | |
1377 | cluster); | |
1378 | ra_index += max_cluster; | |
1379 | } | |
940100a4 | 1380 | |
ecb8bea8 | 1381 | mutex_lock(&inode->i_mutex); |
633085c7 FDBM |
1382 | if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) |
1383 | BTRFS_I(inode)->force_compress = compress_type; | |
008873ea | 1384 | ret = cluster_pages_for_defrag(inode, pages, i, cluster); |
ecb8bea8 LB |
1385 | if (ret < 0) { |
1386 | mutex_unlock(&inode->i_mutex); | |
4cb5300b | 1387 | goto out_ra; |
ecb8bea8 | 1388 | } |
4cb5300b CM |
1389 | |
1390 | defrag_count += ret; | |
d0e1d66b | 1391 | balance_dirty_pages_ratelimited(inode->i_mapping); |
ecb8bea8 | 1392 | mutex_unlock(&inode->i_mutex); |
4cb5300b CM |
1393 | |
1394 | if (newer_than) { | |
1395 | if (newer_off == (u64)-1) | |
1396 | break; | |
1397 | ||
e1f041e1 LB |
1398 | if (ret > 0) |
1399 | i += ret; | |
1400 | ||
4cb5300b CM |
1401 | newer_off = max(newer_off + 1, |
1402 | (u64)i << PAGE_CACHE_SHIFT); | |
1403 | ||
1404 | ret = find_new_extents(root, inode, | |
1405 | newer_than, &newer_off, | |
1406 | 64 * 1024); | |
1407 | if (!ret) { | |
1408 | range->start = newer_off; | |
1409 | i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; | |
4cb5300b CM |
1410 | } else { |
1411 | break; | |
f46b5a66 | 1412 | } |
4cb5300b | 1413 | } else { |
008873ea | 1414 | if (ret > 0) { |
cbcc8326 | 1415 | i += ret; |
008873ea LZ |
1416 | last_len += ret << PAGE_CACHE_SHIFT; |
1417 | } else { | |
cbcc8326 | 1418 | i++; |
008873ea LZ |
1419 | last_len = 0; |
1420 | } | |
f46b5a66 | 1421 | } |
f46b5a66 CH |
1422 | } |
1423 | ||
dec8ef90 | 1424 | if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { |
1e701a32 | 1425 | filemap_flush(inode->i_mapping); |
dec8ef90 FM |
1426 | if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, |
1427 | &BTRFS_I(inode)->runtime_flags)) | |
1428 | filemap_flush(inode->i_mapping); | |
1429 | } | |
1e701a32 CM |
1430 | |
1431 | if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { | |
1432 | /* the filemap_flush will queue IO into the worker threads, but | |
1433 | * we have to make sure the IO is actually started and that | |
1434 | * ordered extents get created before we return | |
1435 | */ | |
1436 | atomic_inc(&root->fs_info->async_submit_draining); | |
1437 | while (atomic_read(&root->fs_info->nr_async_submits) || | |
1438 | atomic_read(&root->fs_info->async_delalloc_pages)) { | |
1439 | wait_event(root->fs_info->async_submit_wait, | |
1440 | (atomic_read(&root->fs_info->nr_async_submits) == 0 && | |
1441 | atomic_read(&root->fs_info->async_delalloc_pages) == 0)); | |
1442 | } | |
1443 | atomic_dec(&root->fs_info->async_submit_draining); | |
1e701a32 CM |
1444 | } |
1445 | ||
1a419d85 | 1446 | if (range->compress_type == BTRFS_COMPRESS_LZO) { |
2b0ce2c2 | 1447 | btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO); |
1a419d85 LZ |
1448 | } |
1449 | ||
60ccf82f | 1450 | ret = defrag_count; |
940100a4 | 1451 | |
4cb5300b | 1452 | out_ra: |
633085c7 FDBM |
1453 | if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { |
1454 | mutex_lock(&inode->i_mutex); | |
1455 | BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE; | |
1456 | mutex_unlock(&inode->i_mutex); | |
1457 | } | |
4cb5300b CM |
1458 | if (!file) |
1459 | kfree(ra); | |
1460 | kfree(pages); | |
940100a4 | 1461 | return ret; |
f46b5a66 CH |
1462 | } |
1463 | ||
198605a8 | 1464 | static noinline int btrfs_ioctl_resize(struct file *file, |
76dda93c | 1465 | void __user *arg) |
f46b5a66 CH |
1466 | { |
1467 | u64 new_size; | |
1468 | u64 old_size; | |
1469 | u64 devid = 1; | |
496ad9aa | 1470 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
f46b5a66 CH |
1471 | struct btrfs_ioctl_vol_args *vol_args; |
1472 | struct btrfs_trans_handle *trans; | |
1473 | struct btrfs_device *device = NULL; | |
1474 | char *sizestr; | |
1475 | char *devstr = NULL; | |
1476 | int ret = 0; | |
f46b5a66 CH |
1477 | int mod = 0; |
1478 | ||
e441d54d CM |
1479 | if (!capable(CAP_SYS_ADMIN)) |
1480 | return -EPERM; | |
1481 | ||
198605a8 MX |
1482 | ret = mnt_want_write_file(file); |
1483 | if (ret) | |
1484 | return ret; | |
1485 | ||
5ac00add SB |
1486 | if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, |
1487 | 1)) { | |
97547676 | 1488 | mnt_drop_write_file(file); |
e57138b3 | 1489 | return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
c9e9f97b ID |
1490 | } |
1491 | ||
5ac00add | 1492 | mutex_lock(&root->fs_info->volume_mutex); |
dae7b665 | 1493 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
1494 | if (IS_ERR(vol_args)) { |
1495 | ret = PTR_ERR(vol_args); | |
1496 | goto out; | |
1497 | } | |
5516e595 MF |
1498 | |
1499 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
f46b5a66 | 1500 | |
f46b5a66 CH |
1501 | sizestr = vol_args->name; |
1502 | devstr = strchr(sizestr, ':'); | |
1503 | if (devstr) { | |
1504 | char *end; | |
1505 | sizestr = devstr + 1; | |
1506 | *devstr = '\0'; | |
1507 | devstr = vol_args->name; | |
1508 | devid = simple_strtoull(devstr, &end, 10); | |
dfd79829 MX |
1509 | if (!devid) { |
1510 | ret = -EINVAL; | |
1511 | goto out_free; | |
1512 | } | |
efe120a0 | 1513 | btrfs_info(root->fs_info, "resizing devid %llu", devid); |
f46b5a66 | 1514 | } |
dba60f3f | 1515 | |
aa1b8cd4 | 1516 | device = btrfs_find_device(root->fs_info, devid, NULL, NULL); |
f46b5a66 | 1517 | if (!device) { |
efe120a0 | 1518 | btrfs_info(root->fs_info, "resizer unable to find device %llu", |
c1c9ff7c | 1519 | devid); |
dfd79829 | 1520 | ret = -ENODEV; |
c9e9f97b | 1521 | goto out_free; |
f46b5a66 | 1522 | } |
dba60f3f MX |
1523 | |
1524 | if (!device->writeable) { | |
efe120a0 FH |
1525 | btrfs_info(root->fs_info, |
1526 | "resizer unable to apply on readonly device %llu", | |
c1c9ff7c | 1527 | devid); |
dfd79829 | 1528 | ret = -EPERM; |
4e42ae1b LB |
1529 | goto out_free; |
1530 | } | |
1531 | ||
f46b5a66 CH |
1532 | if (!strcmp(sizestr, "max")) |
1533 | new_size = device->bdev->bd_inode->i_size; | |
1534 | else { | |
1535 | if (sizestr[0] == '-') { | |
1536 | mod = -1; | |
1537 | sizestr++; | |
1538 | } else if (sizestr[0] == '+') { | |
1539 | mod = 1; | |
1540 | sizestr++; | |
1541 | } | |
91748467 | 1542 | new_size = memparse(sizestr, NULL); |
f46b5a66 CH |
1543 | if (new_size == 0) { |
1544 | ret = -EINVAL; | |
c9e9f97b | 1545 | goto out_free; |
f46b5a66 CH |
1546 | } |
1547 | } | |
1548 | ||
63a212ab | 1549 | if (device->is_tgtdev_for_dev_replace) { |
dfd79829 | 1550 | ret = -EPERM; |
63a212ab SB |
1551 | goto out_free; |
1552 | } | |
1553 | ||
f46b5a66 CH |
1554 | old_size = device->total_bytes; |
1555 | ||
1556 | if (mod < 0) { | |
1557 | if (new_size > old_size) { | |
1558 | ret = -EINVAL; | |
c9e9f97b | 1559 | goto out_free; |
f46b5a66 CH |
1560 | } |
1561 | new_size = old_size - new_size; | |
1562 | } else if (mod > 0) { | |
eb8052e0 WF |
1563 | if (new_size > ULLONG_MAX - old_size) { |
1564 | ret = -EINVAL; | |
1565 | goto out_free; | |
1566 | } | |
f46b5a66 CH |
1567 | new_size = old_size + new_size; |
1568 | } | |
1569 | ||
1570 | if (new_size < 256 * 1024 * 1024) { | |
1571 | ret = -EINVAL; | |
c9e9f97b | 1572 | goto out_free; |
f46b5a66 CH |
1573 | } |
1574 | if (new_size > device->bdev->bd_inode->i_size) { | |
1575 | ret = -EFBIG; | |
c9e9f97b | 1576 | goto out_free; |
f46b5a66 CH |
1577 | } |
1578 | ||
1579 | do_div(new_size, root->sectorsize); | |
1580 | new_size *= root->sectorsize; | |
1581 | ||
efe120a0 | 1582 | printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n", |
c1c9ff7c | 1583 | rcu_str_deref(device->name), new_size); |
f46b5a66 CH |
1584 | |
1585 | if (new_size > old_size) { | |
a22285a6 | 1586 | trans = btrfs_start_transaction(root, 0); |
98d5dc13 TI |
1587 | if (IS_ERR(trans)) { |
1588 | ret = PTR_ERR(trans); | |
c9e9f97b | 1589 | goto out_free; |
98d5dc13 | 1590 | } |
f46b5a66 CH |
1591 | ret = btrfs_grow_device(trans, device, new_size); |
1592 | btrfs_commit_transaction(trans, root); | |
ece7d20e | 1593 | } else if (new_size < old_size) { |
f46b5a66 | 1594 | ret = btrfs_shrink_device(device, new_size); |
0253f40e | 1595 | } /* equal, nothing need to do */ |
f46b5a66 | 1596 | |
c9e9f97b | 1597 | out_free: |
f46b5a66 | 1598 | kfree(vol_args); |
c9e9f97b ID |
1599 | out: |
1600 | mutex_unlock(&root->fs_info->volume_mutex); | |
5ac00add | 1601 | atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); |
18f39c41 | 1602 | mnt_drop_write_file(file); |
f46b5a66 CH |
1603 | return ret; |
1604 | } | |
1605 | ||
72fd032e | 1606 | static noinline int btrfs_ioctl_snap_create_transid(struct file *file, |
6f72c7e2 AJ |
1607 | char *name, unsigned long fd, int subvol, |
1608 | u64 *transid, bool readonly, | |
8696c533 | 1609 | struct btrfs_qgroup_inherit *inherit) |
f46b5a66 | 1610 | { |
f46b5a66 | 1611 | int namelen; |
3de4586c | 1612 | int ret = 0; |
f46b5a66 | 1613 | |
a874a63e LB |
1614 | ret = mnt_want_write_file(file); |
1615 | if (ret) | |
1616 | goto out; | |
1617 | ||
72fd032e SW |
1618 | namelen = strlen(name); |
1619 | if (strchr(name, '/')) { | |
f46b5a66 | 1620 | ret = -EINVAL; |
a874a63e | 1621 | goto out_drop_write; |
f46b5a66 CH |
1622 | } |
1623 | ||
16780cab CM |
1624 | if (name[0] == '.' && |
1625 | (namelen == 1 || (name[1] == '.' && namelen == 2))) { | |
1626 | ret = -EEXIST; | |
a874a63e | 1627 | goto out_drop_write; |
16780cab CM |
1628 | } |
1629 | ||
3de4586c | 1630 | if (subvol) { |
72fd032e | 1631 | ret = btrfs_mksubvol(&file->f_path, name, namelen, |
6f72c7e2 | 1632 | NULL, transid, readonly, inherit); |
cb8e7090 | 1633 | } else { |
2903ff01 | 1634 | struct fd src = fdget(fd); |
3de4586c | 1635 | struct inode *src_inode; |
2903ff01 | 1636 | if (!src.file) { |
3de4586c | 1637 | ret = -EINVAL; |
a874a63e | 1638 | goto out_drop_write; |
3de4586c CM |
1639 | } |
1640 | ||
496ad9aa AV |
1641 | src_inode = file_inode(src.file); |
1642 | if (src_inode->i_sb != file_inode(file)->i_sb) { | |
efe120a0 FH |
1643 | btrfs_info(BTRFS_I(src_inode)->root->fs_info, |
1644 | "Snapshot src from another FS"); | |
23ad5b17 | 1645 | ret = -EXDEV; |
d0242061 DS |
1646 | } else if (!inode_owner_or_capable(src_inode)) { |
1647 | /* | |
1648 | * Subvolume creation is not restricted, but snapshots | |
1649 | * are limited to own subvolumes only | |
1650 | */ | |
1651 | ret = -EPERM; | |
ecd18815 AV |
1652 | } else { |
1653 | ret = btrfs_mksubvol(&file->f_path, name, namelen, | |
1654 | BTRFS_I(src_inode)->root, | |
1655 | transid, readonly, inherit); | |
3de4586c | 1656 | } |
2903ff01 | 1657 | fdput(src); |
cb8e7090 | 1658 | } |
a874a63e LB |
1659 | out_drop_write: |
1660 | mnt_drop_write_file(file); | |
f46b5a66 | 1661 | out: |
72fd032e SW |
1662 | return ret; |
1663 | } | |
1664 | ||
1665 | static noinline int btrfs_ioctl_snap_create(struct file *file, | |
fa0d2b9b | 1666 | void __user *arg, int subvol) |
72fd032e | 1667 | { |
fa0d2b9b | 1668 | struct btrfs_ioctl_vol_args *vol_args; |
72fd032e SW |
1669 | int ret; |
1670 | ||
fa0d2b9b LZ |
1671 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1672 | if (IS_ERR(vol_args)) | |
1673 | return PTR_ERR(vol_args); | |
1674 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
72fd032e | 1675 | |
fa0d2b9b | 1676 | ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, |
b83cc969 | 1677 | vol_args->fd, subvol, |
6f72c7e2 | 1678 | NULL, false, NULL); |
fdfb1e4f | 1679 | |
fa0d2b9b LZ |
1680 | kfree(vol_args); |
1681 | return ret; | |
1682 | } | |
fdfb1e4f | 1683 | |
fa0d2b9b LZ |
1684 | static noinline int btrfs_ioctl_snap_create_v2(struct file *file, |
1685 | void __user *arg, int subvol) | |
1686 | { | |
1687 | struct btrfs_ioctl_vol_args_v2 *vol_args; | |
1688 | int ret; | |
1689 | u64 transid = 0; | |
1690 | u64 *ptr = NULL; | |
b83cc969 | 1691 | bool readonly = false; |
6f72c7e2 | 1692 | struct btrfs_qgroup_inherit *inherit = NULL; |
75eaa0e2 | 1693 | |
fa0d2b9b LZ |
1694 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1695 | if (IS_ERR(vol_args)) | |
1696 | return PTR_ERR(vol_args); | |
1697 | vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; | |
75eaa0e2 | 1698 | |
b83cc969 | 1699 | if (vol_args->flags & |
6f72c7e2 AJ |
1700 | ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | |
1701 | BTRFS_SUBVOL_QGROUP_INHERIT)) { | |
b83cc969 | 1702 | ret = -EOPNOTSUPP; |
fa0d2b9b | 1703 | goto out; |
72fd032e | 1704 | } |
fa0d2b9b LZ |
1705 | |
1706 | if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) | |
1707 | ptr = &transid; | |
b83cc969 LZ |
1708 | if (vol_args->flags & BTRFS_SUBVOL_RDONLY) |
1709 | readonly = true; | |
6f72c7e2 AJ |
1710 | if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { |
1711 | if (vol_args->size > PAGE_CACHE_SIZE) { | |
1712 | ret = -EINVAL; | |
1713 | goto out; | |
1714 | } | |
1715 | inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); | |
1716 | if (IS_ERR(inherit)) { | |
1717 | ret = PTR_ERR(inherit); | |
1718 | goto out; | |
1719 | } | |
1720 | } | |
fa0d2b9b LZ |
1721 | |
1722 | ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, | |
6f72c7e2 | 1723 | vol_args->fd, subvol, ptr, |
8696c533 | 1724 | readonly, inherit); |
fa0d2b9b LZ |
1725 | |
1726 | if (ret == 0 && ptr && | |
1727 | copy_to_user(arg + | |
1728 | offsetof(struct btrfs_ioctl_vol_args_v2, | |
1729 | transid), ptr, sizeof(*ptr))) | |
1730 | ret = -EFAULT; | |
fdfb1e4f | 1731 | out: |
f46b5a66 | 1732 | kfree(vol_args); |
6f72c7e2 | 1733 | kfree(inherit); |
f46b5a66 CH |
1734 | return ret; |
1735 | } | |
1736 | ||
0caa102d LZ |
1737 | static noinline int btrfs_ioctl_subvol_getflags(struct file *file, |
1738 | void __user *arg) | |
1739 | { | |
496ad9aa | 1740 | struct inode *inode = file_inode(file); |
0caa102d LZ |
1741 | struct btrfs_root *root = BTRFS_I(inode)->root; |
1742 | int ret = 0; | |
1743 | u64 flags = 0; | |
1744 | ||
33345d01 | 1745 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) |
0caa102d LZ |
1746 | return -EINVAL; |
1747 | ||
1748 | down_read(&root->fs_info->subvol_sem); | |
1749 | if (btrfs_root_readonly(root)) | |
1750 | flags |= BTRFS_SUBVOL_RDONLY; | |
1751 | up_read(&root->fs_info->subvol_sem); | |
1752 | ||
1753 | if (copy_to_user(arg, &flags, sizeof(flags))) | |
1754 | ret = -EFAULT; | |
1755 | ||
1756 | return ret; | |
1757 | } | |
1758 | ||
1759 | static noinline int btrfs_ioctl_subvol_setflags(struct file *file, | |
1760 | void __user *arg) | |
1761 | { | |
496ad9aa | 1762 | struct inode *inode = file_inode(file); |
0caa102d LZ |
1763 | struct btrfs_root *root = BTRFS_I(inode)->root; |
1764 | struct btrfs_trans_handle *trans; | |
1765 | u64 root_flags; | |
1766 | u64 flags; | |
1767 | int ret = 0; | |
1768 | ||
bd60ea0f DS |
1769 | if (!inode_owner_or_capable(inode)) |
1770 | return -EPERM; | |
1771 | ||
b9ca0664 LB |
1772 | ret = mnt_want_write_file(file); |
1773 | if (ret) | |
1774 | goto out; | |
0caa102d | 1775 | |
b9ca0664 LB |
1776 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { |
1777 | ret = -EINVAL; | |
1778 | goto out_drop_write; | |
1779 | } | |
0caa102d | 1780 | |
b9ca0664 LB |
1781 | if (copy_from_user(&flags, arg, sizeof(flags))) { |
1782 | ret = -EFAULT; | |
1783 | goto out_drop_write; | |
1784 | } | |
0caa102d | 1785 | |
b9ca0664 LB |
1786 | if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { |
1787 | ret = -EINVAL; | |
1788 | goto out_drop_write; | |
1789 | } | |
0caa102d | 1790 | |
b9ca0664 LB |
1791 | if (flags & ~BTRFS_SUBVOL_RDONLY) { |
1792 | ret = -EOPNOTSUPP; | |
1793 | goto out_drop_write; | |
1794 | } | |
0caa102d LZ |
1795 | |
1796 | down_write(&root->fs_info->subvol_sem); | |
1797 | ||
1798 | /* nothing to do */ | |
1799 | if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) | |
b9ca0664 | 1800 | goto out_drop_sem; |
0caa102d LZ |
1801 | |
1802 | root_flags = btrfs_root_flags(&root->root_item); | |
2c686537 | 1803 | if (flags & BTRFS_SUBVOL_RDONLY) { |
0caa102d LZ |
1804 | btrfs_set_root_flags(&root->root_item, |
1805 | root_flags | BTRFS_ROOT_SUBVOL_RDONLY); | |
2c686537 DS |
1806 | } else { |
1807 | /* | |
1808 | * Block RO -> RW transition if this subvolume is involved in | |
1809 | * send | |
1810 | */ | |
1811 | spin_lock(&root->root_item_lock); | |
1812 | if (root->send_in_progress == 0) { | |
1813 | btrfs_set_root_flags(&root->root_item, | |
0caa102d | 1814 | root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); |
2c686537 DS |
1815 | spin_unlock(&root->root_item_lock); |
1816 | } else { | |
1817 | spin_unlock(&root->root_item_lock); | |
1818 | btrfs_warn(root->fs_info, | |
1819 | "Attempt to set subvolume %llu read-write during send", | |
1820 | root->root_key.objectid); | |
1821 | ret = -EPERM; | |
1822 | goto out_drop_sem; | |
1823 | } | |
1824 | } | |
0caa102d LZ |
1825 | |
1826 | trans = btrfs_start_transaction(root, 1); | |
1827 | if (IS_ERR(trans)) { | |
1828 | ret = PTR_ERR(trans); | |
1829 | goto out_reset; | |
1830 | } | |
1831 | ||
b4dc2b8c | 1832 | ret = btrfs_update_root(trans, root->fs_info->tree_root, |
0caa102d LZ |
1833 | &root->root_key, &root->root_item); |
1834 | ||
1835 | btrfs_commit_transaction(trans, root); | |
1836 | out_reset: | |
1837 | if (ret) | |
1838 | btrfs_set_root_flags(&root->root_item, root_flags); | |
b9ca0664 | 1839 | out_drop_sem: |
0caa102d | 1840 | up_write(&root->fs_info->subvol_sem); |
b9ca0664 LB |
1841 | out_drop_write: |
1842 | mnt_drop_write_file(file); | |
1843 | out: | |
0caa102d LZ |
1844 | return ret; |
1845 | } | |
1846 | ||
76dda93c YZ |
1847 | /* |
1848 | * helper to check if the subvolume references other subvolumes | |
1849 | */ | |
1850 | static noinline int may_destroy_subvol(struct btrfs_root *root) | |
1851 | { | |
1852 | struct btrfs_path *path; | |
175a2b87 | 1853 | struct btrfs_dir_item *di; |
76dda93c | 1854 | struct btrfs_key key; |
175a2b87 | 1855 | u64 dir_id; |
76dda93c YZ |
1856 | int ret; |
1857 | ||
1858 | path = btrfs_alloc_path(); | |
1859 | if (!path) | |
1860 | return -ENOMEM; | |
1861 | ||
175a2b87 JB |
1862 | /* Make sure this root isn't set as the default subvol */ |
1863 | dir_id = btrfs_super_root_dir(root->fs_info->super_copy); | |
1864 | di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path, | |
1865 | dir_id, "default", 7, 0); | |
1866 | if (di && !IS_ERR(di)) { | |
1867 | btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key); | |
1868 | if (key.objectid == root->root_key.objectid) { | |
72de6b53 GS |
1869 | ret = -EPERM; |
1870 | btrfs_err(root->fs_info, "deleting default subvolume " | |
1871 | "%llu is not allowed", key.objectid); | |
175a2b87 JB |
1872 | goto out; |
1873 | } | |
1874 | btrfs_release_path(path); | |
1875 | } | |
1876 | ||
76dda93c YZ |
1877 | key.objectid = root->root_key.objectid; |
1878 | key.type = BTRFS_ROOT_REF_KEY; | |
1879 | key.offset = (u64)-1; | |
1880 | ||
1881 | ret = btrfs_search_slot(NULL, root->fs_info->tree_root, | |
1882 | &key, path, 0, 0); | |
1883 | if (ret < 0) | |
1884 | goto out; | |
1885 | BUG_ON(ret == 0); | |
1886 | ||
1887 | ret = 0; | |
1888 | if (path->slots[0] > 0) { | |
1889 | path->slots[0]--; | |
1890 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); | |
1891 | if (key.objectid == root->root_key.objectid && | |
1892 | key.type == BTRFS_ROOT_REF_KEY) | |
1893 | ret = -ENOTEMPTY; | |
1894 | } | |
1895 | out: | |
1896 | btrfs_free_path(path); | |
1897 | return ret; | |
1898 | } | |
1899 | ||
ac8e9819 CM |
1900 | static noinline int key_in_sk(struct btrfs_key *key, |
1901 | struct btrfs_ioctl_search_key *sk) | |
1902 | { | |
abc6e134 CM |
1903 | struct btrfs_key test; |
1904 | int ret; | |
1905 | ||
1906 | test.objectid = sk->min_objectid; | |
1907 | test.type = sk->min_type; | |
1908 | test.offset = sk->min_offset; | |
1909 | ||
1910 | ret = btrfs_comp_cpu_keys(key, &test); | |
1911 | if (ret < 0) | |
ac8e9819 | 1912 | return 0; |
abc6e134 CM |
1913 | |
1914 | test.objectid = sk->max_objectid; | |
1915 | test.type = sk->max_type; | |
1916 | test.offset = sk->max_offset; | |
1917 | ||
1918 | ret = btrfs_comp_cpu_keys(key, &test); | |
1919 | if (ret > 0) | |
ac8e9819 CM |
1920 | return 0; |
1921 | return 1; | |
1922 | } | |
1923 | ||
1924 | static noinline int copy_to_sk(struct btrfs_root *root, | |
1925 | struct btrfs_path *path, | |
1926 | struct btrfs_key *key, | |
1927 | struct btrfs_ioctl_search_key *sk, | |
1928 | char *buf, | |
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 | ||
1960 | if (sizeof(sh) + item_len > BTRFS_SEARCH_ARGS_BUFSIZE) | |
ac8e9819 CM |
1961 | item_len = 0; |
1962 | ||
1963 | if (sizeof(sh) + item_len + *sk_offset > | |
1964 | BTRFS_SEARCH_ARGS_BUFSIZE) { | |
1965 | ret = 1; | |
1966 | goto overflow; | |
1967 | } | |
1968 | ||
ac8e9819 CM |
1969 | sh.objectid = key->objectid; |
1970 | sh.offset = key->offset; | |
1971 | sh.type = key->type; | |
1972 | sh.len = item_len; | |
1973 | sh.transid = found_transid; | |
1974 | ||
1975 | /* copy search result header */ | |
1976 | memcpy(buf + *sk_offset, &sh, sizeof(sh)); | |
1977 | *sk_offset += sizeof(sh); | |
1978 | ||
1979 | if (item_len) { | |
1980 | char *p = buf + *sk_offset; | |
1981 | /* copy the item */ | |
1982 | read_extent_buffer(leaf, p, | |
1983 | item_off, item_len); | |
1984 | *sk_offset += item_len; | |
ac8e9819 | 1985 | } |
e2156867 | 1986 | (*num_found)++; |
ac8e9819 CM |
1987 | |
1988 | if (*num_found >= sk->nr_items) | |
1989 | break; | |
1990 | } | |
1991 | advance_key: | |
abc6e134 CM |
1992 | ret = 0; |
1993 | if (key->offset < (u64)-1 && key->offset < sk->max_offset) | |
ac8e9819 | 1994 | key->offset++; |
abc6e134 CM |
1995 | else if (key->type < (u8)-1 && key->type < sk->max_type) { |
1996 | key->offset = 0; | |
ac8e9819 | 1997 | key->type++; |
abc6e134 CM |
1998 | } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) { |
1999 | key->offset = 0; | |
2000 | key->type = 0; | |
ac8e9819 | 2001 | key->objectid++; |
abc6e134 CM |
2002 | } else |
2003 | ret = 1; | |
ac8e9819 | 2004 | overflow: |
ac8e9819 CM |
2005 | return ret; |
2006 | } | |
2007 | ||
2008 | static noinline int search_ioctl(struct inode *inode, | |
2009 | struct btrfs_ioctl_search_args *args) | |
2010 | { | |
2011 | struct btrfs_root *root; | |
2012 | struct btrfs_key key; | |
ac8e9819 CM |
2013 | struct btrfs_path *path; |
2014 | struct btrfs_ioctl_search_key *sk = &args->key; | |
2015 | struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info; | |
2016 | int ret; | |
2017 | int num_found = 0; | |
2018 | unsigned long sk_offset = 0; | |
2019 | ||
2020 | path = btrfs_alloc_path(); | |
2021 | if (!path) | |
2022 | return -ENOMEM; | |
2023 | ||
2024 | if (sk->tree_id == 0) { | |
2025 | /* search the root of the inode that was passed */ | |
2026 | root = BTRFS_I(inode)->root; | |
2027 | } else { | |
2028 | key.objectid = sk->tree_id; | |
2029 | key.type = BTRFS_ROOT_ITEM_KEY; | |
2030 | key.offset = (u64)-1; | |
2031 | root = btrfs_read_fs_root_no_name(info, &key); | |
2032 | if (IS_ERR(root)) { | |
efe120a0 | 2033 | printk(KERN_ERR "BTRFS: could not find root %llu\n", |
ac8e9819 CM |
2034 | sk->tree_id); |
2035 | btrfs_free_path(path); | |
2036 | return -ENOENT; | |
2037 | } | |
2038 | } | |
2039 | ||
2040 | key.objectid = sk->min_objectid; | |
2041 | key.type = sk->min_type; | |
2042 | key.offset = sk->min_offset; | |
2043 | ||
ac8e9819 CM |
2044 | path->keep_locks = 1; |
2045 | ||
67871254 | 2046 | while (1) { |
6174d3cb | 2047 | ret = btrfs_search_forward(root, &key, path, sk->min_transid); |
ac8e9819 CM |
2048 | if (ret != 0) { |
2049 | if (ret > 0) | |
2050 | ret = 0; | |
2051 | goto err; | |
2052 | } | |
2053 | ret = copy_to_sk(root, path, &key, sk, args->buf, | |
2054 | &sk_offset, &num_found); | |
b3b4aa74 | 2055 | btrfs_release_path(path); |
ac8e9819 CM |
2056 | if (ret || num_found >= sk->nr_items) |
2057 | break; | |
2058 | ||
2059 | } | |
2060 | ret = 0; | |
2061 | err: | |
2062 | sk->nr_items = num_found; | |
2063 | btrfs_free_path(path); | |
2064 | return ret; | |
2065 | } | |
2066 | ||
2067 | static noinline int btrfs_ioctl_tree_search(struct file *file, | |
2068 | void __user *argp) | |
2069 | { | |
2070 | struct btrfs_ioctl_search_args *args; | |
2071 | struct inode *inode; | |
2072 | int ret; | |
2073 | ||
2074 | if (!capable(CAP_SYS_ADMIN)) | |
2075 | return -EPERM; | |
2076 | ||
2354d08f JL |
2077 | args = memdup_user(argp, sizeof(*args)); |
2078 | if (IS_ERR(args)) | |
2079 | return PTR_ERR(args); | |
ac8e9819 | 2080 | |
496ad9aa | 2081 | inode = file_inode(file); |
ac8e9819 CM |
2082 | ret = search_ioctl(inode, args); |
2083 | if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) | |
2084 | ret = -EFAULT; | |
2085 | kfree(args); | |
2086 | return ret; | |
2087 | } | |
2088 | ||
98d377a0 | 2089 | /* |
ac8e9819 CM |
2090 | * Search INODE_REFs to identify path name of 'dirid' directory |
2091 | * in a 'tree_id' tree. and sets path name to 'name'. | |
2092 | */ | |
98d377a0 TH |
2093 | static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, |
2094 | u64 tree_id, u64 dirid, char *name) | |
2095 | { | |
2096 | struct btrfs_root *root; | |
2097 | struct btrfs_key key; | |
ac8e9819 | 2098 | char *ptr; |
98d377a0 TH |
2099 | int ret = -1; |
2100 | int slot; | |
2101 | int len; | |
2102 | int total_len = 0; | |
2103 | struct btrfs_inode_ref *iref; | |
2104 | struct extent_buffer *l; | |
2105 | struct btrfs_path *path; | |
2106 | ||
2107 | if (dirid == BTRFS_FIRST_FREE_OBJECTID) { | |
2108 | name[0]='\0'; | |
2109 | return 0; | |
2110 | } | |
2111 | ||
2112 | path = btrfs_alloc_path(); | |
2113 | if (!path) | |
2114 | return -ENOMEM; | |
2115 | ||
ac8e9819 | 2116 | ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; |
98d377a0 TH |
2117 | |
2118 | key.objectid = tree_id; | |
2119 | key.type = BTRFS_ROOT_ITEM_KEY; | |
2120 | key.offset = (u64)-1; | |
2121 | root = btrfs_read_fs_root_no_name(info, &key); | |
2122 | if (IS_ERR(root)) { | |
efe120a0 | 2123 | printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id); |
8ad6fcab CM |
2124 | ret = -ENOENT; |
2125 | goto out; | |
98d377a0 TH |
2126 | } |
2127 | ||
2128 | key.objectid = dirid; | |
2129 | key.type = BTRFS_INODE_REF_KEY; | |
8ad6fcab | 2130 | key.offset = (u64)-1; |
98d377a0 | 2131 | |
67871254 | 2132 | while (1) { |
98d377a0 TH |
2133 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
2134 | if (ret < 0) | |
2135 | goto out; | |
18674c6c FDBM |
2136 | else if (ret > 0) { |
2137 | ret = btrfs_previous_item(root, path, dirid, | |
2138 | BTRFS_INODE_REF_KEY); | |
2139 | if (ret < 0) | |
2140 | goto out; | |
2141 | else if (ret > 0) { | |
2142 | ret = -ENOENT; | |
2143 | goto out; | |
2144 | } | |
2145 | } | |
98d377a0 TH |
2146 | |
2147 | l = path->nodes[0]; | |
2148 | slot = path->slots[0]; | |
2149 | btrfs_item_key_to_cpu(l, &key, slot); | |
2150 | ||
98d377a0 TH |
2151 | iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); |
2152 | len = btrfs_inode_ref_name_len(l, iref); | |
2153 | ptr -= len + 1; | |
2154 | total_len += len + 1; | |
a696cf35 FDBM |
2155 | if (ptr < name) { |
2156 | ret = -ENAMETOOLONG; | |
98d377a0 | 2157 | goto out; |
a696cf35 | 2158 | } |
98d377a0 TH |
2159 | |
2160 | *(ptr + len) = '/'; | |
67871254 | 2161 | read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); |
98d377a0 TH |
2162 | |
2163 | if (key.offset == BTRFS_FIRST_FREE_OBJECTID) | |
2164 | break; | |
2165 | ||
b3b4aa74 | 2166 | btrfs_release_path(path); |
98d377a0 | 2167 | key.objectid = key.offset; |
8ad6fcab | 2168 | key.offset = (u64)-1; |
98d377a0 | 2169 | dirid = key.objectid; |
98d377a0 | 2170 | } |
77906a50 | 2171 | memmove(name, ptr, total_len); |
67871254 | 2172 | name[total_len] = '\0'; |
98d377a0 TH |
2173 | ret = 0; |
2174 | out: | |
2175 | btrfs_free_path(path); | |
ac8e9819 CM |
2176 | return ret; |
2177 | } | |
2178 | ||
2179 | static noinline int btrfs_ioctl_ino_lookup(struct file *file, | |
2180 | void __user *argp) | |
2181 | { | |
2182 | struct btrfs_ioctl_ino_lookup_args *args; | |
2183 | struct inode *inode; | |
2184 | int ret; | |
2185 | ||
2186 | if (!capable(CAP_SYS_ADMIN)) | |
2187 | return -EPERM; | |
2188 | ||
2354d08f JL |
2189 | args = memdup_user(argp, sizeof(*args)); |
2190 | if (IS_ERR(args)) | |
2191 | return PTR_ERR(args); | |
c2b96929 | 2192 | |
496ad9aa | 2193 | inode = file_inode(file); |
ac8e9819 | 2194 | |
1b53ac4d CM |
2195 | if (args->treeid == 0) |
2196 | args->treeid = BTRFS_I(inode)->root->root_key.objectid; | |
2197 | ||
ac8e9819 CM |
2198 | ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, |
2199 | args->treeid, args->objectid, | |
2200 | args->name); | |
2201 | ||
2202 | if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) | |
2203 | ret = -EFAULT; | |
2204 | ||
2205 | kfree(args); | |
98d377a0 TH |
2206 | return ret; |
2207 | } | |
2208 | ||
76dda93c YZ |
2209 | static noinline int btrfs_ioctl_snap_destroy(struct file *file, |
2210 | void __user *arg) | |
2211 | { | |
54563d41 | 2212 | struct dentry *parent = file->f_path.dentry; |
76dda93c YZ |
2213 | struct dentry *dentry; |
2214 | struct inode *dir = parent->d_inode; | |
2215 | struct inode *inode; | |
2216 | struct btrfs_root *root = BTRFS_I(dir)->root; | |
2217 | struct btrfs_root *dest = NULL; | |
2218 | struct btrfs_ioctl_vol_args *vol_args; | |
2219 | struct btrfs_trans_handle *trans; | |
c58aaad2 MX |
2220 | struct btrfs_block_rsv block_rsv; |
2221 | u64 qgroup_reserved; | |
76dda93c YZ |
2222 | int namelen; |
2223 | int ret; | |
2224 | int err = 0; | |
2225 | ||
76dda93c YZ |
2226 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
2227 | if (IS_ERR(vol_args)) | |
2228 | return PTR_ERR(vol_args); | |
2229 | ||
2230 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
2231 | namelen = strlen(vol_args->name); | |
2232 | if (strchr(vol_args->name, '/') || | |
2233 | strncmp(vol_args->name, "..", namelen) == 0) { | |
2234 | err = -EINVAL; | |
2235 | goto out; | |
2236 | } | |
2237 | ||
a561be71 | 2238 | err = mnt_want_write_file(file); |
76dda93c YZ |
2239 | if (err) |
2240 | goto out; | |
2241 | ||
5c50c9b8 DS |
2242 | err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT); |
2243 | if (err == -EINTR) | |
e43f998e | 2244 | goto out_drop_write; |
76dda93c YZ |
2245 | dentry = lookup_one_len(vol_args->name, parent, namelen); |
2246 | if (IS_ERR(dentry)) { | |
2247 | err = PTR_ERR(dentry); | |
2248 | goto out_unlock_dir; | |
2249 | } | |
2250 | ||
2251 | if (!dentry->d_inode) { | |
2252 | err = -ENOENT; | |
2253 | goto out_dput; | |
2254 | } | |
2255 | ||
2256 | inode = dentry->d_inode; | |
4260f7c7 | 2257 | dest = BTRFS_I(inode)->root; |
67871254 | 2258 | if (!capable(CAP_SYS_ADMIN)) { |
4260f7c7 SW |
2259 | /* |
2260 | * Regular user. Only allow this with a special mount | |
2261 | * option, when the user has write+exec access to the | |
2262 | * subvol root, and when rmdir(2) would have been | |
2263 | * allowed. | |
2264 | * | |
2265 | * Note that this is _not_ check that the subvol is | |
2266 | * empty or doesn't contain data that we wouldn't | |
2267 | * otherwise be able to delete. | |
2268 | * | |
2269 | * Users who want to delete empty subvols should try | |
2270 | * rmdir(2). | |
2271 | */ | |
2272 | err = -EPERM; | |
2273 | if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED)) | |
2274 | goto out_dput; | |
2275 | ||
2276 | /* | |
2277 | * Do not allow deletion if the parent dir is the same | |
2278 | * as the dir to be deleted. That means the ioctl | |
2279 | * must be called on the dentry referencing the root | |
2280 | * of the subvol, not a random directory contained | |
2281 | * within it. | |
2282 | */ | |
2283 | err = -EINVAL; | |
2284 | if (root == dest) | |
2285 | goto out_dput; | |
2286 | ||
2287 | err = inode_permission(inode, MAY_WRITE | MAY_EXEC); | |
2288 | if (err) | |
2289 | goto out_dput; | |
4260f7c7 SW |
2290 | } |
2291 | ||
5c39da5b MX |
2292 | /* check if subvolume may be deleted by a user */ |
2293 | err = btrfs_may_delete(dir, dentry, 1); | |
2294 | if (err) | |
2295 | goto out_dput; | |
2296 | ||
33345d01 | 2297 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { |
76dda93c YZ |
2298 | err = -EINVAL; |
2299 | goto out_dput; | |
2300 | } | |
2301 | ||
76dda93c YZ |
2302 | mutex_lock(&inode->i_mutex); |
2303 | err = d_invalidate(dentry); | |
2304 | if (err) | |
2305 | goto out_unlock; | |
2306 | ||
2307 | down_write(&root->fs_info->subvol_sem); | |
2308 | ||
2309 | err = may_destroy_subvol(dest); | |
2310 | if (err) | |
2311 | goto out_up_write; | |
2312 | ||
c58aaad2 MX |
2313 | btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); |
2314 | /* | |
2315 | * One for dir inode, two for dir entries, two for root | |
2316 | * ref/backref. | |
2317 | */ | |
2318 | err = btrfs_subvolume_reserve_metadata(root, &block_rsv, | |
ee3441b4 | 2319 | 5, &qgroup_reserved, true); |
c58aaad2 MX |
2320 | if (err) |
2321 | goto out_up_write; | |
2322 | ||
a22285a6 YZ |
2323 | trans = btrfs_start_transaction(root, 0); |
2324 | if (IS_ERR(trans)) { | |
2325 | err = PTR_ERR(trans); | |
c58aaad2 | 2326 | goto out_release; |
a22285a6 | 2327 | } |
c58aaad2 MX |
2328 | trans->block_rsv = &block_rsv; |
2329 | trans->bytes_reserved = block_rsv.size; | |
a22285a6 | 2330 | |
76dda93c YZ |
2331 | ret = btrfs_unlink_subvol(trans, root, dir, |
2332 | dest->root_key.objectid, | |
2333 | dentry->d_name.name, | |
2334 | dentry->d_name.len); | |
79787eaa JM |
2335 | if (ret) { |
2336 | err = ret; | |
2337 | btrfs_abort_transaction(trans, root, ret); | |
2338 | goto out_end_trans; | |
2339 | } | |
76dda93c YZ |
2340 | |
2341 | btrfs_record_root_in_trans(trans, dest); | |
2342 | ||
2343 | memset(&dest->root_item.drop_progress, 0, | |
2344 | sizeof(dest->root_item.drop_progress)); | |
2345 | dest->root_item.drop_level = 0; | |
2346 | btrfs_set_root_refs(&dest->root_item, 0); | |
2347 | ||
d68fc57b YZ |
2348 | if (!xchg(&dest->orphan_item_inserted, 1)) { |
2349 | ret = btrfs_insert_orphan_item(trans, | |
2350 | root->fs_info->tree_root, | |
2351 | dest->root_key.objectid); | |
79787eaa JM |
2352 | if (ret) { |
2353 | btrfs_abort_transaction(trans, root, ret); | |
2354 | err = ret; | |
2355 | goto out_end_trans; | |
2356 | } | |
d68fc57b | 2357 | } |
dd5f9615 SB |
2358 | |
2359 | ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, | |
2360 | dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL, | |
2361 | dest->root_key.objectid); | |
2362 | if (ret && ret != -ENOENT) { | |
2363 | btrfs_abort_transaction(trans, root, ret); | |
2364 | err = ret; | |
2365 | goto out_end_trans; | |
2366 | } | |
2367 | if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) { | |
2368 | ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, | |
2369 | dest->root_item.received_uuid, | |
2370 | BTRFS_UUID_KEY_RECEIVED_SUBVOL, | |
2371 | dest->root_key.objectid); | |
2372 | if (ret && ret != -ENOENT) { | |
2373 | btrfs_abort_transaction(trans, root, ret); | |
2374 | err = ret; | |
2375 | goto out_end_trans; | |
2376 | } | |
2377 | } | |
2378 | ||
79787eaa | 2379 | out_end_trans: |
c58aaad2 MX |
2380 | trans->block_rsv = NULL; |
2381 | trans->bytes_reserved = 0; | |
531cb13f | 2382 | ret = btrfs_end_transaction(trans, root); |
79787eaa JM |
2383 | if (ret && !err) |
2384 | err = ret; | |
76dda93c | 2385 | inode->i_flags |= S_DEAD; |
c58aaad2 MX |
2386 | out_release: |
2387 | btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); | |
76dda93c YZ |
2388 | out_up_write: |
2389 | up_write(&root->fs_info->subvol_sem); | |
2390 | out_unlock: | |
2391 | mutex_unlock(&inode->i_mutex); | |
2392 | if (!err) { | |
efefb143 | 2393 | shrink_dcache_sb(root->fs_info->sb); |
76dda93c YZ |
2394 | btrfs_invalidate_inodes(dest); |
2395 | d_delete(dentry); | |
fa6ac876 LB |
2396 | |
2397 | /* the last ref */ | |
2398 | if (dest->cache_inode) { | |
2399 | iput(dest->cache_inode); | |
2400 | dest->cache_inode = NULL; | |
2401 | } | |
76dda93c YZ |
2402 | } |
2403 | out_dput: | |
2404 | dput(dentry); | |
2405 | out_unlock_dir: | |
2406 | mutex_unlock(&dir->i_mutex); | |
e43f998e | 2407 | out_drop_write: |
2a79f17e | 2408 | mnt_drop_write_file(file); |
76dda93c YZ |
2409 | out: |
2410 | kfree(vol_args); | |
2411 | return err; | |
2412 | } | |
2413 | ||
1e701a32 | 2414 | static int btrfs_ioctl_defrag(struct file *file, void __user *argp) |
f46b5a66 | 2415 | { |
496ad9aa | 2416 | struct inode *inode = file_inode(file); |
f46b5a66 | 2417 | struct btrfs_root *root = BTRFS_I(inode)->root; |
1e701a32 | 2418 | struct btrfs_ioctl_defrag_range_args *range; |
c146afad YZ |
2419 | int ret; |
2420 | ||
25122d15 ID |
2421 | ret = mnt_want_write_file(file); |
2422 | if (ret) | |
2423 | return ret; | |
b83cc969 | 2424 | |
25122d15 ID |
2425 | if (btrfs_root_readonly(root)) { |
2426 | ret = -EROFS; | |
2427 | goto out; | |
5ac00add | 2428 | } |
f46b5a66 CH |
2429 | |
2430 | switch (inode->i_mode & S_IFMT) { | |
2431 | case S_IFDIR: | |
e441d54d CM |
2432 | if (!capable(CAP_SYS_ADMIN)) { |
2433 | ret = -EPERM; | |
2434 | goto out; | |
2435 | } | |
de78b51a | 2436 | ret = btrfs_defrag_root(root); |
8929ecfa YZ |
2437 | if (ret) |
2438 | goto out; | |
de78b51a | 2439 | ret = btrfs_defrag_root(root->fs_info->extent_root); |
f46b5a66 CH |
2440 | break; |
2441 | case S_IFREG: | |
e441d54d CM |
2442 | if (!(file->f_mode & FMODE_WRITE)) { |
2443 | ret = -EINVAL; | |
2444 | goto out; | |
2445 | } | |
1e701a32 CM |
2446 | |
2447 | range = kzalloc(sizeof(*range), GFP_KERNEL); | |
2448 | if (!range) { | |
2449 | ret = -ENOMEM; | |
2450 | goto out; | |
2451 | } | |
2452 | ||
2453 | if (argp) { | |
2454 | if (copy_from_user(range, argp, | |
2455 | sizeof(*range))) { | |
2456 | ret = -EFAULT; | |
2457 | kfree(range); | |
683be16e | 2458 | goto out; |
1e701a32 CM |
2459 | } |
2460 | /* compression requires us to start the IO */ | |
2461 | if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { | |
2462 | range->flags |= BTRFS_DEFRAG_RANGE_START_IO; | |
2463 | range->extent_thresh = (u32)-1; | |
2464 | } | |
2465 | } else { | |
2466 | /* the rest are all set to zero by kzalloc */ | |
2467 | range->len = (u64)-1; | |
2468 | } | |
496ad9aa | 2469 | ret = btrfs_defrag_file(file_inode(file), file, |
4cb5300b CM |
2470 | range, 0, 0); |
2471 | if (ret > 0) | |
2472 | ret = 0; | |
1e701a32 | 2473 | kfree(range); |
f46b5a66 | 2474 | break; |
8929ecfa YZ |
2475 | default: |
2476 | ret = -EINVAL; | |
f46b5a66 | 2477 | } |
e441d54d | 2478 | out: |
25122d15 | 2479 | mnt_drop_write_file(file); |
e441d54d | 2480 | return ret; |
f46b5a66 CH |
2481 | } |
2482 | ||
b2950863 | 2483 | static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) |
f46b5a66 CH |
2484 | { |
2485 | struct btrfs_ioctl_vol_args *vol_args; | |
2486 | int ret; | |
2487 | ||
e441d54d CM |
2488 | if (!capable(CAP_SYS_ADMIN)) |
2489 | return -EPERM; | |
2490 | ||
5ac00add SB |
2491 | if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, |
2492 | 1)) { | |
e57138b3 | 2493 | return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
c9e9f97b ID |
2494 | } |
2495 | ||
5ac00add | 2496 | mutex_lock(&root->fs_info->volume_mutex); |
dae7b665 | 2497 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
2498 | if (IS_ERR(vol_args)) { |
2499 | ret = PTR_ERR(vol_args); | |
2500 | goto out; | |
2501 | } | |
f46b5a66 | 2502 | |
5516e595 | 2503 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
f46b5a66 CH |
2504 | ret = btrfs_init_new_device(root, vol_args->name); |
2505 | ||
f46b5a66 | 2506 | kfree(vol_args); |
c9e9f97b ID |
2507 | out: |
2508 | mutex_unlock(&root->fs_info->volume_mutex); | |
5ac00add | 2509 | atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); |
f46b5a66 CH |
2510 | return ret; |
2511 | } | |
2512 | ||
da24927b | 2513 | static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) |
f46b5a66 | 2514 | { |
496ad9aa | 2515 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
f46b5a66 CH |
2516 | struct btrfs_ioctl_vol_args *vol_args; |
2517 | int ret; | |
2518 | ||
e441d54d CM |
2519 | if (!capable(CAP_SYS_ADMIN)) |
2520 | return -EPERM; | |
2521 | ||
da24927b MX |
2522 | ret = mnt_want_write_file(file); |
2523 | if (ret) | |
2524 | return ret; | |
c146afad | 2525 | |
dae7b665 | 2526 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
2527 | if (IS_ERR(vol_args)) { |
2528 | ret = PTR_ERR(vol_args); | |
2529 | goto out; | |
2530 | } | |
f46b5a66 | 2531 | |
5516e595 | 2532 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
f46b5a66 | 2533 | |
183860f6 AJ |
2534 | if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, |
2535 | 1)) { | |
2536 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; | |
2537 | goto out; | |
2538 | } | |
2539 | ||
2540 | mutex_lock(&root->fs_info->volume_mutex); | |
2541 | ret = btrfs_rm_device(root, vol_args->name); | |
c9e9f97b | 2542 | mutex_unlock(&root->fs_info->volume_mutex); |
5ac00add | 2543 | atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); |
183860f6 AJ |
2544 | |
2545 | out: | |
2546 | kfree(vol_args); | |
4ac20c70 | 2547 | mnt_drop_write_file(file); |
f46b5a66 CH |
2548 | return ret; |
2549 | } | |
2550 | ||
475f6387 JS |
2551 | static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg) |
2552 | { | |
027ed2f0 | 2553 | struct btrfs_ioctl_fs_info_args *fi_args; |
475f6387 JS |
2554 | struct btrfs_device *device; |
2555 | struct btrfs_device *next; | |
2556 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
027ed2f0 | 2557 | int ret = 0; |
475f6387 JS |
2558 | |
2559 | if (!capable(CAP_SYS_ADMIN)) | |
2560 | return -EPERM; | |
2561 | ||
027ed2f0 LZ |
2562 | fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); |
2563 | if (!fi_args) | |
2564 | return -ENOMEM; | |
2565 | ||
f7171750 | 2566 | mutex_lock(&fs_devices->device_list_mutex); |
027ed2f0 LZ |
2567 | fi_args->num_devices = fs_devices->num_devices; |
2568 | memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid)); | |
475f6387 | 2569 | |
475f6387 | 2570 | list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) { |
027ed2f0 LZ |
2571 | if (device->devid > fi_args->max_id) |
2572 | fi_args->max_id = device->devid; | |
475f6387 JS |
2573 | } |
2574 | mutex_unlock(&fs_devices->device_list_mutex); | |
2575 | ||
027ed2f0 LZ |
2576 | if (copy_to_user(arg, fi_args, sizeof(*fi_args))) |
2577 | ret = -EFAULT; | |
475f6387 | 2578 | |
027ed2f0 LZ |
2579 | kfree(fi_args); |
2580 | return ret; | |
475f6387 JS |
2581 | } |
2582 | ||
2583 | static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg) | |
2584 | { | |
2585 | struct btrfs_ioctl_dev_info_args *di_args; | |
2586 | struct btrfs_device *dev; | |
2587 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
2588 | int ret = 0; | |
2589 | char *s_uuid = NULL; | |
475f6387 JS |
2590 | |
2591 | if (!capable(CAP_SYS_ADMIN)) | |
2592 | return -EPERM; | |
2593 | ||
2594 | di_args = memdup_user(arg, sizeof(*di_args)); | |
2595 | if (IS_ERR(di_args)) | |
2596 | return PTR_ERR(di_args); | |
2597 | ||
dd5f9615 | 2598 | if (!btrfs_is_empty_uuid(di_args->uuid)) |
475f6387 JS |
2599 | s_uuid = di_args->uuid; |
2600 | ||
2601 | mutex_lock(&fs_devices->device_list_mutex); | |
aa1b8cd4 | 2602 | dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL); |
475f6387 JS |
2603 | |
2604 | if (!dev) { | |
2605 | ret = -ENODEV; | |
2606 | goto out; | |
2607 | } | |
2608 | ||
2609 | di_args->devid = dev->devid; | |
2610 | di_args->bytes_used = dev->bytes_used; | |
2611 | di_args->total_bytes = dev->total_bytes; | |
2612 | memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); | |
a27202fb | 2613 | if (dev->name) { |
606686ee JB |
2614 | struct rcu_string *name; |
2615 | ||
2616 | rcu_read_lock(); | |
2617 | name = rcu_dereference(dev->name); | |
2618 | strncpy(di_args->path, name->str, sizeof(di_args->path)); | |
2619 | rcu_read_unlock(); | |
a27202fb JM |
2620 | di_args->path[sizeof(di_args->path) - 1] = 0; |
2621 | } else { | |
99ba55ad | 2622 | di_args->path[0] = '\0'; |
a27202fb | 2623 | } |
475f6387 JS |
2624 | |
2625 | out: | |
55793c0d | 2626 | mutex_unlock(&fs_devices->device_list_mutex); |
475f6387 JS |
2627 | if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) |
2628 | ret = -EFAULT; | |
2629 | ||
2630 | kfree(di_args); | |
2631 | return ret; | |
2632 | } | |
2633 | ||
416161db MF |
2634 | static struct page *extent_same_get_page(struct inode *inode, u64 off) |
2635 | { | |
2636 | struct page *page; | |
2637 | pgoff_t index; | |
2638 | struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree; | |
2639 | ||
2640 | index = off >> PAGE_CACHE_SHIFT; | |
2641 | ||
2642 | page = grab_cache_page(inode->i_mapping, index); | |
2643 | if (!page) | |
2644 | return NULL; | |
2645 | ||
2646 | if (!PageUptodate(page)) { | |
2647 | if (extent_read_full_page_nolock(tree, page, btrfs_get_extent, | |
2648 | 0)) | |
2649 | return NULL; | |
2650 | lock_page(page); | |
2651 | if (!PageUptodate(page)) { | |
2652 | unlock_page(page); | |
2653 | page_cache_release(page); | |
2654 | return NULL; | |
2655 | } | |
2656 | } | |
2657 | unlock_page(page); | |
2658 | ||
2659 | return page; | |
2660 | } | |
2661 | ||
77fe20dc MF |
2662 | static inline void lock_extent_range(struct inode *inode, u64 off, u64 len) |
2663 | { | |
2664 | /* do any pending delalloc/csum calc on src, one way or | |
2665 | another, and lock file content */ | |
2666 | while (1) { | |
2667 | struct btrfs_ordered_extent *ordered; | |
2668 | lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); | |
2669 | ordered = btrfs_lookup_first_ordered_extent(inode, | |
2670 | off + len - 1); | |
2671 | if (!ordered && | |
2672 | !test_range_bit(&BTRFS_I(inode)->io_tree, off, | |
2673 | off + len - 1, EXTENT_DELALLOC, 0, NULL)) | |
2674 | break; | |
2675 | unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); | |
2676 | if (ordered) | |
2677 | btrfs_put_ordered_extent(ordered); | |
2678 | btrfs_wait_ordered_range(inode, off, len); | |
2679 | } | |
2680 | } | |
2681 | ||
416161db MF |
2682 | static void btrfs_double_unlock(struct inode *inode1, u64 loff1, |
2683 | struct inode *inode2, u64 loff2, u64 len) | |
2684 | { | |
2685 | unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); | |
2686 | unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); | |
2687 | ||
2688 | mutex_unlock(&inode1->i_mutex); | |
2689 | mutex_unlock(&inode2->i_mutex); | |
2690 | } | |
2691 | ||
2692 | static void btrfs_double_lock(struct inode *inode1, u64 loff1, | |
2693 | struct inode *inode2, u64 loff2, u64 len) | |
2694 | { | |
2695 | if (inode1 < inode2) { | |
2696 | swap(inode1, inode2); | |
2697 | swap(loff1, loff2); | |
2698 | } | |
2699 | ||
2700 | mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT); | |
2701 | lock_extent_range(inode1, loff1, len); | |
2702 | if (inode1 != inode2) { | |
2703 | mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD); | |
2704 | lock_extent_range(inode2, loff2, len); | |
2705 | } | |
2706 | } | |
2707 | ||
2708 | static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst, | |
2709 | u64 dst_loff, u64 len) | |
2710 | { | |
2711 | int ret = 0; | |
2712 | struct page *src_page, *dst_page; | |
2713 | unsigned int cmp_len = PAGE_CACHE_SIZE; | |
2714 | void *addr, *dst_addr; | |
2715 | ||
2716 | while (len) { | |
2717 | if (len < PAGE_CACHE_SIZE) | |
2718 | cmp_len = len; | |
2719 | ||
2720 | src_page = extent_same_get_page(src, loff); | |
2721 | if (!src_page) | |
2722 | return -EINVAL; | |
2723 | dst_page = extent_same_get_page(dst, dst_loff); | |
2724 | if (!dst_page) { | |
2725 | page_cache_release(src_page); | |
2726 | return -EINVAL; | |
2727 | } | |
2728 | addr = kmap_atomic(src_page); | |
2729 | dst_addr = kmap_atomic(dst_page); | |
2730 | ||
2731 | flush_dcache_page(src_page); | |
2732 | flush_dcache_page(dst_page); | |
2733 | ||
2734 | if (memcmp(addr, dst_addr, cmp_len)) | |
2735 | ret = BTRFS_SAME_DATA_DIFFERS; | |
2736 | ||
2737 | kunmap_atomic(addr); | |
2738 | kunmap_atomic(dst_addr); | |
2739 | page_cache_release(src_page); | |
2740 | page_cache_release(dst_page); | |
2741 | ||
2742 | if (ret) | |
2743 | break; | |
2744 | ||
2745 | loff += cmp_len; | |
2746 | dst_loff += cmp_len; | |
2747 | len -= cmp_len; | |
2748 | } | |
2749 | ||
2750 | return ret; | |
2751 | } | |
2752 | ||
2753 | static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len) | |
2754 | { | |
2755 | u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; | |
2756 | ||
2757 | if (off + len > inode->i_size || off + len < off) | |
2758 | return -EINVAL; | |
2759 | /* Check that we are block aligned - btrfs_clone() requires this */ | |
2760 | if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) | |
2761 | return -EINVAL; | |
2762 | ||
2763 | return 0; | |
2764 | } | |
2765 | ||
2766 | static int btrfs_extent_same(struct inode *src, u64 loff, u64 len, | |
2767 | struct inode *dst, u64 dst_loff) | |
2768 | { | |
2769 | int ret; | |
2770 | ||
2771 | /* | |
2772 | * btrfs_clone() can't handle extents in the same file | |
2773 | * yet. Once that works, we can drop this check and replace it | |
2774 | * with a check for the same inode, but overlapping extents. | |
2775 | */ | |
2776 | if (src == dst) | |
2777 | return -EINVAL; | |
2778 | ||
2779 | btrfs_double_lock(src, loff, dst, dst_loff, len); | |
2780 | ||
2781 | ret = extent_same_check_offsets(src, loff, len); | |
2782 | if (ret) | |
2783 | goto out_unlock; | |
2784 | ||
2785 | ret = extent_same_check_offsets(dst, dst_loff, len); | |
2786 | if (ret) | |
2787 | goto out_unlock; | |
2788 | ||
2789 | /* don't make the dst file partly checksummed */ | |
2790 | if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != | |
2791 | (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { | |
2792 | ret = -EINVAL; | |
2793 | goto out_unlock; | |
2794 | } | |
2795 | ||
2796 | ret = btrfs_cmp_data(src, loff, dst, dst_loff, len); | |
2797 | if (ret == 0) | |
2798 | ret = btrfs_clone(src, dst, loff, len, len, dst_loff); | |
2799 | ||
2800 | out_unlock: | |
2801 | btrfs_double_unlock(src, loff, dst, dst_loff, len); | |
2802 | ||
2803 | return ret; | |
2804 | } | |
2805 | ||
2806 | #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024) | |
2807 | ||
2808 | static long btrfs_ioctl_file_extent_same(struct file *file, | |
2809 | void __user *argp) | |
2810 | { | |
cbf8b8ca MF |
2811 | struct btrfs_ioctl_same_args tmp; |
2812 | struct btrfs_ioctl_same_args *same; | |
2813 | struct btrfs_ioctl_same_extent_info *info; | |
416161db MF |
2814 | struct inode *src = file->f_dentry->d_inode; |
2815 | struct file *dst_file = NULL; | |
2816 | struct inode *dst; | |
2817 | u64 off; | |
2818 | u64 len; | |
2819 | int i; | |
2820 | int ret; | |
cbf8b8ca | 2821 | unsigned long size; |
416161db MF |
2822 | u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; |
2823 | bool is_admin = capable(CAP_SYS_ADMIN); | |
2824 | ||
2825 | if (!(file->f_mode & FMODE_READ)) | |
2826 | return -EINVAL; | |
2827 | ||
2828 | ret = mnt_want_write_file(file); | |
2829 | if (ret) | |
2830 | return ret; | |
2831 | ||
cbf8b8ca | 2832 | if (copy_from_user(&tmp, |
416161db | 2833 | (struct btrfs_ioctl_same_args __user *)argp, |
cbf8b8ca | 2834 | sizeof(tmp))) { |
416161db MF |
2835 | ret = -EFAULT; |
2836 | goto out; | |
2837 | } | |
2838 | ||
cbf8b8ca MF |
2839 | size = sizeof(tmp) + |
2840 | tmp.dest_count * sizeof(struct btrfs_ioctl_same_extent_info); | |
2841 | ||
229eed43 | 2842 | same = memdup_user((struct btrfs_ioctl_same_args __user *)argp, size); |
cbf8b8ca | 2843 | |
229eed43 GB |
2844 | if (IS_ERR(same)) { |
2845 | ret = PTR_ERR(same); | |
cbf8b8ca MF |
2846 | goto out; |
2847 | } | |
2848 | ||
2849 | off = same->logical_offset; | |
2850 | len = same->length; | |
416161db MF |
2851 | |
2852 | /* | |
2853 | * Limit the total length we will dedupe for each operation. | |
2854 | * This is intended to bound the total time spent in this | |
2855 | * ioctl to something sane. | |
2856 | */ | |
2857 | if (len > BTRFS_MAX_DEDUPE_LEN) | |
2858 | len = BTRFS_MAX_DEDUPE_LEN; | |
2859 | ||
2860 | if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) { | |
2861 | /* | |
2862 | * Btrfs does not support blocksize < page_size. As a | |
2863 | * result, btrfs_cmp_data() won't correctly handle | |
2864 | * this situation without an update. | |
2865 | */ | |
2866 | ret = -EINVAL; | |
2867 | goto out; | |
2868 | } | |
2869 | ||
2870 | ret = -EISDIR; | |
2871 | if (S_ISDIR(src->i_mode)) | |
2872 | goto out; | |
2873 | ||
2874 | ret = -EACCES; | |
2875 | if (!S_ISREG(src->i_mode)) | |
2876 | goto out; | |
2877 | ||
cbf8b8ca MF |
2878 | /* pre-format output fields to sane values */ |
2879 | for (i = 0; i < same->dest_count; i++) { | |
2880 | same->info[i].bytes_deduped = 0ULL; | |
2881 | same->info[i].status = 0; | |
2882 | } | |
416161db | 2883 | |
cbf8b8ca MF |
2884 | ret = 0; |
2885 | for (i = 0; i < same->dest_count; i++) { | |
2886 | info = &same->info[i]; | |
416161db | 2887 | |
cbf8b8ca | 2888 | dst_file = fget(info->fd); |
416161db | 2889 | if (!dst_file) { |
cbf8b8ca | 2890 | info->status = -EBADF; |
416161db MF |
2891 | goto next; |
2892 | } | |
2893 | ||
2894 | if (!(is_admin || (dst_file->f_mode & FMODE_WRITE))) { | |
cbf8b8ca | 2895 | info->status = -EINVAL; |
416161db MF |
2896 | goto next; |
2897 | } | |
2898 | ||
cbf8b8ca | 2899 | info->status = -EXDEV; |
416161db MF |
2900 | if (file->f_path.mnt != dst_file->f_path.mnt) |
2901 | goto next; | |
2902 | ||
2903 | dst = dst_file->f_dentry->d_inode; | |
2904 | if (src->i_sb != dst->i_sb) | |
2905 | goto next; | |
2906 | ||
2907 | if (S_ISDIR(dst->i_mode)) { | |
cbf8b8ca | 2908 | info->status = -EISDIR; |
416161db MF |
2909 | goto next; |
2910 | } | |
2911 | ||
2912 | if (!S_ISREG(dst->i_mode)) { | |
cbf8b8ca | 2913 | info->status = -EACCES; |
416161db MF |
2914 | goto next; |
2915 | } | |
2916 | ||
cbf8b8ca MF |
2917 | info->status = btrfs_extent_same(src, off, len, dst, |
2918 | info->logical_offset); | |
2919 | if (info->status == 0) | |
2920 | info->bytes_deduped += len; | |
416161db MF |
2921 | |
2922 | next: | |
2923 | if (dst_file) | |
2924 | fput(dst_file); | |
416161db MF |
2925 | } |
2926 | ||
cbf8b8ca MF |
2927 | ret = copy_to_user(argp, same, size); |
2928 | if (ret) | |
2929 | ret = -EFAULT; | |
2930 | ||
416161db MF |
2931 | out: |
2932 | mnt_drop_write_file(file); | |
2933 | return ret; | |
2934 | } | |
2935 | ||
32b7c687 MF |
2936 | /** |
2937 | * btrfs_clone() - clone a range from inode file to another | |
2938 | * | |
2939 | * @src: Inode to clone from | |
2940 | * @inode: Inode to clone to | |
2941 | * @off: Offset within source to start clone from | |
2942 | * @olen: Original length, passed by user, of range to clone | |
2943 | * @olen_aligned: Block-aligned value of olen, extent_same uses | |
2944 | * identical values here | |
2945 | * @destoff: Offset within @inode to start clone | |
2946 | */ | |
2947 | static int btrfs_clone(struct inode *src, struct inode *inode, | |
2948 | u64 off, u64 olen, u64 olen_aligned, u64 destoff) | |
f46b5a66 | 2949 | { |
f46b5a66 | 2950 | struct btrfs_root *root = BTRFS_I(inode)->root; |
32b7c687 | 2951 | struct btrfs_path *path = NULL; |
f46b5a66 | 2952 | struct extent_buffer *leaf; |
32b7c687 MF |
2953 | struct btrfs_trans_handle *trans; |
2954 | char *buf = NULL; | |
ae01a0ab | 2955 | struct btrfs_key key; |
f46b5a66 CH |
2956 | u32 nritems; |
2957 | int slot; | |
ae01a0ab | 2958 | int ret; |
32b7c687 | 2959 | u64 len = olen_aligned; |
ae01a0ab YZ |
2960 | |
2961 | ret = -ENOMEM; | |
2962 | buf = vmalloc(btrfs_level_size(root, 0)); | |
2963 | if (!buf) | |
32b7c687 | 2964 | return ret; |
ae01a0ab YZ |
2965 | |
2966 | path = btrfs_alloc_path(); | |
2967 | if (!path) { | |
2968 | vfree(buf); | |
32b7c687 | 2969 | return ret; |
d525e8ab LZ |
2970 | } |
2971 | ||
32b7c687 | 2972 | path->reada = 2; |
c5c9cd4d | 2973 | /* clone data */ |
33345d01 | 2974 | key.objectid = btrfs_ino(src); |
ae01a0ab YZ |
2975 | key.type = BTRFS_EXTENT_DATA_KEY; |
2976 | key.offset = 0; | |
f46b5a66 CH |
2977 | |
2978 | while (1) { | |
2979 | /* | |
2980 | * note the key will change type as we walk through the | |
2981 | * tree. | |
2982 | */ | |
e4355f34 | 2983 | path->leave_spinning = 1; |
362a20c5 DS |
2984 | ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, |
2985 | 0, 0); | |
f46b5a66 CH |
2986 | if (ret < 0) |
2987 | goto out; | |
2988 | ||
ae01a0ab | 2989 | nritems = btrfs_header_nritems(path->nodes[0]); |
e4355f34 | 2990 | process_slot: |
ae01a0ab | 2991 | if (path->slots[0] >= nritems) { |
362a20c5 | 2992 | ret = btrfs_next_leaf(BTRFS_I(src)->root, path); |
f46b5a66 CH |
2993 | if (ret < 0) |
2994 | goto out; | |
2995 | if (ret > 0) | |
2996 | break; | |
ae01a0ab | 2997 | nritems = btrfs_header_nritems(path->nodes[0]); |
f46b5a66 CH |
2998 | } |
2999 | leaf = path->nodes[0]; | |
3000 | slot = path->slots[0]; | |
f46b5a66 | 3001 | |
ae01a0ab | 3002 | btrfs_item_key_to_cpu(leaf, &key, slot); |
d20f7043 | 3003 | if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY || |
33345d01 | 3004 | key.objectid != btrfs_ino(src)) |
f46b5a66 CH |
3005 | break; |
3006 | ||
c5c9cd4d SW |
3007 | if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { |
3008 | struct btrfs_file_extent_item *extent; | |
3009 | int type; | |
31840ae1 ZY |
3010 | u32 size; |
3011 | struct btrfs_key new_key; | |
c5c9cd4d SW |
3012 | u64 disko = 0, diskl = 0; |
3013 | u64 datao = 0, datal = 0; | |
3014 | u8 comp; | |
b5384d48 | 3015 | u64 endoff; |
31840ae1 | 3016 | |
c5c9cd4d SW |
3017 | extent = btrfs_item_ptr(leaf, slot, |
3018 | struct btrfs_file_extent_item); | |
3019 | comp = btrfs_file_extent_compression(leaf, extent); | |
3020 | type = btrfs_file_extent_type(leaf, extent); | |
c8a894d7 CM |
3021 | if (type == BTRFS_FILE_EXTENT_REG || |
3022 | type == BTRFS_FILE_EXTENT_PREALLOC) { | |
d397712b CM |
3023 | disko = btrfs_file_extent_disk_bytenr(leaf, |
3024 | extent); | |
3025 | diskl = btrfs_file_extent_disk_num_bytes(leaf, | |
3026 | extent); | |
c5c9cd4d | 3027 | datao = btrfs_file_extent_offset(leaf, extent); |
d397712b CM |
3028 | datal = btrfs_file_extent_num_bytes(leaf, |
3029 | extent); | |
c5c9cd4d SW |
3030 | } else if (type == BTRFS_FILE_EXTENT_INLINE) { |
3031 | /* take upper bound, may be compressed */ | |
3032 | datal = btrfs_file_extent_ram_bytes(leaf, | |
3033 | extent); | |
3034 | } | |
31840ae1 | 3035 | |
050006a7 | 3036 | if (key.offset + datal <= off || |
e4355f34 FDBM |
3037 | key.offset >= off + len - 1) { |
3038 | path->slots[0]++; | |
3039 | goto process_slot; | |
3040 | } | |
3041 | ||
3042 | size = btrfs_item_size_nr(leaf, slot); | |
3043 | read_extent_buffer(leaf, buf, | |
3044 | btrfs_item_ptr_offset(leaf, slot), | |
3045 | size); | |
3046 | ||
3047 | btrfs_release_path(path); | |
3048 | path->leave_spinning = 0; | |
c5c9cd4d | 3049 | |
31840ae1 | 3050 | memcpy(&new_key, &key, sizeof(new_key)); |
33345d01 | 3051 | new_key.objectid = btrfs_ino(inode); |
4d728ec7 LZ |
3052 | if (off <= key.offset) |
3053 | new_key.offset = key.offset + destoff - off; | |
3054 | else | |
3055 | new_key.offset = destoff; | |
31840ae1 | 3056 | |
b6f3409b SW |
3057 | /* |
3058 | * 1 - adjusting old extent (we may have to split it) | |
3059 | * 1 - add new extent | |
3060 | * 1 - inode update | |
3061 | */ | |
3062 | trans = btrfs_start_transaction(root, 3); | |
a22285a6 YZ |
3063 | if (IS_ERR(trans)) { |
3064 | ret = PTR_ERR(trans); | |
3065 | goto out; | |
3066 | } | |
3067 | ||
c8a894d7 CM |
3068 | if (type == BTRFS_FILE_EXTENT_REG || |
3069 | type == BTRFS_FILE_EXTENT_PREALLOC) { | |
d72c0842 LZ |
3070 | /* |
3071 | * a | --- range to clone ---| b | |
3072 | * | ------------- extent ------------- | | |
3073 | */ | |
3074 | ||
3075 | /* substract range b */ | |
3076 | if (key.offset + datal > off + len) | |
3077 | datal = off + len - key.offset; | |
3078 | ||
3079 | /* substract range a */ | |
a22285a6 YZ |
3080 | if (off > key.offset) { |
3081 | datao += off - key.offset; | |
3082 | datal -= off - key.offset; | |
3083 | } | |
3084 | ||
5dc562c5 | 3085 | ret = btrfs_drop_extents(trans, root, inode, |
a22285a6 YZ |
3086 | new_key.offset, |
3087 | new_key.offset + datal, | |
2671485d | 3088 | 1); |
79787eaa JM |
3089 | if (ret) { |
3090 | btrfs_abort_transaction(trans, root, | |
3091 | ret); | |
3092 | btrfs_end_transaction(trans, root); | |
3093 | goto out; | |
3094 | } | |
a22285a6 | 3095 | |
c5c9cd4d SW |
3096 | ret = btrfs_insert_empty_item(trans, root, path, |
3097 | &new_key, size); | |
79787eaa JM |
3098 | if (ret) { |
3099 | btrfs_abort_transaction(trans, root, | |
3100 | ret); | |
3101 | btrfs_end_transaction(trans, root); | |
3102 | goto out; | |
3103 | } | |
c5c9cd4d SW |
3104 | |
3105 | leaf = path->nodes[0]; | |
3106 | slot = path->slots[0]; | |
3107 | write_extent_buffer(leaf, buf, | |
31840ae1 ZY |
3108 | btrfs_item_ptr_offset(leaf, slot), |
3109 | size); | |
ae01a0ab | 3110 | |
c5c9cd4d | 3111 | extent = btrfs_item_ptr(leaf, slot, |
f46b5a66 | 3112 | struct btrfs_file_extent_item); |
c5c9cd4d | 3113 | |
c5c9cd4d SW |
3114 | /* disko == 0 means it's a hole */ |
3115 | if (!disko) | |
3116 | datao = 0; | |
c5c9cd4d SW |
3117 | |
3118 | btrfs_set_file_extent_offset(leaf, extent, | |
3119 | datao); | |
3120 | btrfs_set_file_extent_num_bytes(leaf, extent, | |
3121 | datal); | |
3122 | if (disko) { | |
3123 | inode_add_bytes(inode, datal); | |
ae01a0ab | 3124 | ret = btrfs_inc_extent_ref(trans, root, |
5d4f98a2 YZ |
3125 | disko, diskl, 0, |
3126 | root->root_key.objectid, | |
33345d01 | 3127 | btrfs_ino(inode), |
66d7e7f0 AJ |
3128 | new_key.offset - datao, |
3129 | 0); | |
79787eaa JM |
3130 | if (ret) { |
3131 | btrfs_abort_transaction(trans, | |
3132 | root, | |
3133 | ret); | |
3134 | btrfs_end_transaction(trans, | |
3135 | root); | |
3136 | goto out; | |
3137 | ||
3138 | } | |
f46b5a66 | 3139 | } |
c5c9cd4d SW |
3140 | } else if (type == BTRFS_FILE_EXTENT_INLINE) { |
3141 | u64 skip = 0; | |
3142 | u64 trim = 0; | |
3143 | if (off > key.offset) { | |
3144 | skip = off - key.offset; | |
3145 | new_key.offset += skip; | |
3146 | } | |
d397712b | 3147 | |
aa42ffd9 LB |
3148 | if (key.offset + datal > off + len) |
3149 | trim = key.offset + datal - (off + len); | |
d397712b | 3150 | |
c5c9cd4d | 3151 | if (comp && (skip || trim)) { |
c5c9cd4d | 3152 | ret = -EINVAL; |
a22285a6 | 3153 | btrfs_end_transaction(trans, root); |
c5c9cd4d SW |
3154 | goto out; |
3155 | } | |
3156 | size -= skip + trim; | |
3157 | datal -= skip + trim; | |
a22285a6 | 3158 | |
5dc562c5 | 3159 | ret = btrfs_drop_extents(trans, root, inode, |
a22285a6 YZ |
3160 | new_key.offset, |
3161 | new_key.offset + datal, | |
2671485d | 3162 | 1); |
79787eaa JM |
3163 | if (ret) { |
3164 | btrfs_abort_transaction(trans, root, | |
3165 | ret); | |
3166 | btrfs_end_transaction(trans, root); | |
3167 | goto out; | |
3168 | } | |
a22285a6 | 3169 | |
c5c9cd4d SW |
3170 | ret = btrfs_insert_empty_item(trans, root, path, |
3171 | &new_key, size); | |
79787eaa JM |
3172 | if (ret) { |
3173 | btrfs_abort_transaction(trans, root, | |
3174 | ret); | |
3175 | btrfs_end_transaction(trans, root); | |
3176 | goto out; | |
3177 | } | |
c5c9cd4d SW |
3178 | |
3179 | if (skip) { | |
d397712b CM |
3180 | u32 start = |
3181 | btrfs_file_extent_calc_inline_size(0); | |
c5c9cd4d SW |
3182 | memmove(buf+start, buf+start+skip, |
3183 | datal); | |
3184 | } | |
3185 | ||
3186 | leaf = path->nodes[0]; | |
3187 | slot = path->slots[0]; | |
3188 | write_extent_buffer(leaf, buf, | |
3189 | btrfs_item_ptr_offset(leaf, slot), | |
3190 | size); | |
3191 | inode_add_bytes(inode, datal); | |
f46b5a66 | 3192 | } |
c5c9cd4d SW |
3193 | |
3194 | btrfs_mark_buffer_dirty(leaf); | |
b3b4aa74 | 3195 | btrfs_release_path(path); |
c5c9cd4d | 3196 | |
0c4d2d95 | 3197 | inode_inc_iversion(inode); |
a22285a6 | 3198 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; |
b5384d48 SW |
3199 | |
3200 | /* | |
3201 | * we round up to the block size at eof when | |
3202 | * determining which extents to clone above, | |
3203 | * but shouldn't round up the file size | |
3204 | */ | |
3205 | endoff = new_key.offset + datal; | |
5f3888ff LZ |
3206 | if (endoff > destoff+olen) |
3207 | endoff = destoff+olen; | |
b5384d48 SW |
3208 | if (endoff > inode->i_size) |
3209 | btrfs_i_size_write(inode, endoff); | |
3210 | ||
a22285a6 | 3211 | ret = btrfs_update_inode(trans, root, inode); |
79787eaa JM |
3212 | if (ret) { |
3213 | btrfs_abort_transaction(trans, root, ret); | |
3214 | btrfs_end_transaction(trans, root); | |
3215 | goto out; | |
3216 | } | |
3217 | ret = btrfs_end_transaction(trans, root); | |
a22285a6 | 3218 | } |
b3b4aa74 | 3219 | btrfs_release_path(path); |
f46b5a66 | 3220 | key.offset++; |
f46b5a66 | 3221 | } |
f46b5a66 | 3222 | ret = 0; |
32b7c687 | 3223 | |
f46b5a66 | 3224 | out: |
b3b4aa74 | 3225 | btrfs_release_path(path); |
32b7c687 MF |
3226 | btrfs_free_path(path); |
3227 | vfree(buf); | |
3228 | return ret; | |
3229 | } | |
3230 | ||
3231 | static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, | |
3232 | u64 off, u64 olen, u64 destoff) | |
3233 | { | |
54563d41 | 3234 | struct inode *inode = file_inode(file); |
32b7c687 MF |
3235 | struct btrfs_root *root = BTRFS_I(inode)->root; |
3236 | struct fd src_file; | |
3237 | struct inode *src; | |
3238 | int ret; | |
3239 | u64 len = olen; | |
3240 | u64 bs = root->fs_info->sb->s_blocksize; | |
3241 | int same_inode = 0; | |
3242 | ||
3243 | /* | |
3244 | * TODO: | |
3245 | * - split compressed inline extents. annoying: we need to | |
3246 | * decompress into destination's address_space (the file offset | |
3247 | * may change, so source mapping won't do), then recompress (or | |
3248 | * otherwise reinsert) a subrange. | |
3249 | * - allow ranges within the same file to be cloned (provided | |
3250 | * they don't overlap)? | |
3251 | */ | |
3252 | ||
3253 | /* the destination must be opened for writing */ | |
3254 | if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) | |
3255 | return -EINVAL; | |
3256 | ||
3257 | if (btrfs_root_readonly(root)) | |
3258 | return -EROFS; | |
3259 | ||
3260 | ret = mnt_want_write_file(file); | |
3261 | if (ret) | |
3262 | return ret; | |
3263 | ||
3264 | src_file = fdget(srcfd); | |
3265 | if (!src_file.file) { | |
3266 | ret = -EBADF; | |
3267 | goto out_drop_write; | |
3268 | } | |
3269 | ||
3270 | ret = -EXDEV; | |
3271 | if (src_file.file->f_path.mnt != file->f_path.mnt) | |
3272 | goto out_fput; | |
3273 | ||
3274 | src = file_inode(src_file.file); | |
3275 | ||
3276 | ret = -EINVAL; | |
3277 | if (src == inode) | |
3278 | same_inode = 1; | |
3279 | ||
3280 | /* the src must be open for reading */ | |
3281 | if (!(src_file.file->f_mode & FMODE_READ)) | |
3282 | goto out_fput; | |
3283 | ||
3284 | /* don't make the dst file partly checksummed */ | |
3285 | if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != | |
3286 | (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) | |
3287 | goto out_fput; | |
3288 | ||
3289 | ret = -EISDIR; | |
3290 | if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) | |
3291 | goto out_fput; | |
3292 | ||
3293 | ret = -EXDEV; | |
3294 | if (src->i_sb != inode->i_sb) | |
3295 | goto out_fput; | |
3296 | ||
3297 | if (!same_inode) { | |
3298 | if (inode < src) { | |
3299 | mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT); | |
3300 | mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD); | |
3301 | } else { | |
3302 | mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT); | |
3303 | mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD); | |
3304 | } | |
3305 | } else { | |
3306 | mutex_lock(&src->i_mutex); | |
3307 | } | |
3308 | ||
3309 | /* determine range to clone */ | |
3310 | ret = -EINVAL; | |
3311 | if (off + len > src->i_size || off + len < off) | |
3312 | goto out_unlock; | |
3313 | if (len == 0) | |
3314 | olen = len = src->i_size - off; | |
3315 | /* if we extend to eof, continue to block boundary */ | |
3316 | if (off + len == src->i_size) | |
3317 | len = ALIGN(src->i_size, bs) - off; | |
3318 | ||
3319 | /* verify the end result is block aligned */ | |
3320 | if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || | |
3321 | !IS_ALIGNED(destoff, bs)) | |
3322 | goto out_unlock; | |
3323 | ||
3324 | /* verify if ranges are overlapped within the same file */ | |
3325 | if (same_inode) { | |
3326 | if (destoff + len > off && destoff < off + len) | |
3327 | goto out_unlock; | |
3328 | } | |
3329 | ||
3330 | if (destoff > inode->i_size) { | |
3331 | ret = btrfs_cont_expand(inode, inode->i_size, destoff); | |
3332 | if (ret) | |
3333 | goto out_unlock; | |
3334 | } | |
3335 | ||
3336 | /* truncate page cache pages from target inode range */ | |
3337 | truncate_inode_pages_range(&inode->i_data, destoff, | |
3338 | PAGE_CACHE_ALIGN(destoff + len) - 1); | |
3339 | ||
3340 | lock_extent_range(src, off, len); | |
3341 | ||
3342 | ret = btrfs_clone(src, inode, off, olen, len, destoff); | |
3343 | ||
aa42ffd9 | 3344 | unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1); |
f46b5a66 | 3345 | out_unlock: |
c57c2b3e FDBM |
3346 | if (!same_inode) { |
3347 | if (inode < src) { | |
3348 | mutex_unlock(&src->i_mutex); | |
3349 | mutex_unlock(&inode->i_mutex); | |
3350 | } else { | |
3351 | mutex_unlock(&inode->i_mutex); | |
3352 | mutex_unlock(&src->i_mutex); | |
3353 | } | |
3354 | } else { | |
3355 | mutex_unlock(&src->i_mutex); | |
3356 | } | |
f46b5a66 | 3357 | out_fput: |
2903ff01 | 3358 | fdput(src_file); |
ab67b7c1 | 3359 | out_drop_write: |
2a79f17e | 3360 | mnt_drop_write_file(file); |
f46b5a66 CH |
3361 | return ret; |
3362 | } | |
3363 | ||
7a865e8a | 3364 | static long btrfs_ioctl_clone_range(struct file *file, void __user *argp) |
c5c9cd4d SW |
3365 | { |
3366 | struct btrfs_ioctl_clone_range_args args; | |
3367 | ||
7a865e8a | 3368 | if (copy_from_user(&args, argp, sizeof(args))) |
c5c9cd4d SW |
3369 | return -EFAULT; |
3370 | return btrfs_ioctl_clone(file, args.src_fd, args.src_offset, | |
3371 | args.src_length, args.dest_offset); | |
3372 | } | |
3373 | ||
f46b5a66 CH |
3374 | /* |
3375 | * there are many ways the trans_start and trans_end ioctls can lead | |
3376 | * to deadlocks. They should only be used by applications that | |
3377 | * basically own the machine, and have a very in depth understanding | |
3378 | * of all the possible deadlocks and enospc problems. | |
3379 | */ | |
b2950863 | 3380 | static long btrfs_ioctl_trans_start(struct file *file) |
f46b5a66 | 3381 | { |
496ad9aa | 3382 | struct inode *inode = file_inode(file); |
f46b5a66 CH |
3383 | struct btrfs_root *root = BTRFS_I(inode)->root; |
3384 | struct btrfs_trans_handle *trans; | |
1ab86aed | 3385 | int ret; |
f46b5a66 | 3386 | |
1ab86aed | 3387 | ret = -EPERM; |
df5b5520 | 3388 | if (!capable(CAP_SYS_ADMIN)) |
1ab86aed | 3389 | goto out; |
df5b5520 | 3390 | |
1ab86aed SW |
3391 | ret = -EINPROGRESS; |
3392 | if (file->private_data) | |
f46b5a66 | 3393 | goto out; |
9ca9ee09 | 3394 | |
b83cc969 LZ |
3395 | ret = -EROFS; |
3396 | if (btrfs_root_readonly(root)) | |
3397 | goto out; | |
3398 | ||
a561be71 | 3399 | ret = mnt_want_write_file(file); |
c146afad YZ |
3400 | if (ret) |
3401 | goto out; | |
3402 | ||
a4abeea4 | 3403 | atomic_inc(&root->fs_info->open_ioctl_trans); |
9ca9ee09 | 3404 | |
1ab86aed | 3405 | ret = -ENOMEM; |
7a7eaa40 | 3406 | trans = btrfs_start_ioctl_transaction(root); |
abd30bb0 | 3407 | if (IS_ERR(trans)) |
1ab86aed SW |
3408 | goto out_drop; |
3409 | ||
3410 | file->private_data = trans; | |
3411 | return 0; | |
3412 | ||
3413 | out_drop: | |
a4abeea4 | 3414 | atomic_dec(&root->fs_info->open_ioctl_trans); |
2a79f17e | 3415 | mnt_drop_write_file(file); |
f46b5a66 | 3416 | out: |
f46b5a66 CH |
3417 | return ret; |
3418 | } | |
3419 | ||
6ef5ed0d JB |
3420 | static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) |
3421 | { | |
496ad9aa | 3422 | struct inode *inode = file_inode(file); |
6ef5ed0d JB |
3423 | struct btrfs_root *root = BTRFS_I(inode)->root; |
3424 | struct btrfs_root *new_root; | |
3425 | struct btrfs_dir_item *di; | |
3426 | struct btrfs_trans_handle *trans; | |
3427 | struct btrfs_path *path; | |
3428 | struct btrfs_key location; | |
3429 | struct btrfs_disk_key disk_key; | |
6ef5ed0d JB |
3430 | u64 objectid = 0; |
3431 | u64 dir_id; | |
3c04ce01 | 3432 | int ret; |
6ef5ed0d JB |
3433 | |
3434 | if (!capable(CAP_SYS_ADMIN)) | |
3435 | return -EPERM; | |
3436 | ||
3c04ce01 MX |
3437 | ret = mnt_want_write_file(file); |
3438 | if (ret) | |
3439 | return ret; | |
3440 | ||
3441 | if (copy_from_user(&objectid, argp, sizeof(objectid))) { | |
3442 | ret = -EFAULT; | |
3443 | goto out; | |
3444 | } | |
6ef5ed0d JB |
3445 | |
3446 | if (!objectid) | |
1cecf579 | 3447 | objectid = BTRFS_FS_TREE_OBJECTID; |
6ef5ed0d JB |
3448 | |
3449 | location.objectid = objectid; | |
3450 | location.type = BTRFS_ROOT_ITEM_KEY; | |
3451 | location.offset = (u64)-1; | |
3452 | ||
3453 | new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); | |
3c04ce01 MX |
3454 | if (IS_ERR(new_root)) { |
3455 | ret = PTR_ERR(new_root); | |
3456 | goto out; | |
3457 | } | |
6ef5ed0d | 3458 | |
6ef5ed0d | 3459 | path = btrfs_alloc_path(); |
3c04ce01 MX |
3460 | if (!path) { |
3461 | ret = -ENOMEM; | |
3462 | goto out; | |
3463 | } | |
6ef5ed0d JB |
3464 | path->leave_spinning = 1; |
3465 | ||
3466 | trans = btrfs_start_transaction(root, 1); | |
98d5dc13 | 3467 | if (IS_ERR(trans)) { |
6ef5ed0d | 3468 | btrfs_free_path(path); |
3c04ce01 MX |
3469 | ret = PTR_ERR(trans); |
3470 | goto out; | |
6ef5ed0d JB |
3471 | } |
3472 | ||
6c41761f | 3473 | dir_id = btrfs_super_root_dir(root->fs_info->super_copy); |
6ef5ed0d JB |
3474 | di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path, |
3475 | dir_id, "default", 7, 1); | |
cf1e99a4 | 3476 | if (IS_ERR_OR_NULL(di)) { |
6ef5ed0d JB |
3477 | btrfs_free_path(path); |
3478 | btrfs_end_transaction(trans, root); | |
efe120a0 FH |
3479 | btrfs_err(new_root->fs_info, "Umm, you don't have the default dir" |
3480 | "item, this isn't going to work"); | |
3c04ce01 MX |
3481 | ret = -ENOENT; |
3482 | goto out; | |
6ef5ed0d JB |
3483 | } |
3484 | ||
3485 | btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); | |
3486 | btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); | |
3487 | btrfs_mark_buffer_dirty(path->nodes[0]); | |
3488 | btrfs_free_path(path); | |
3489 | ||
2b0ce2c2 | 3490 | btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL); |
6ef5ed0d | 3491 | btrfs_end_transaction(trans, root); |
3c04ce01 MX |
3492 | out: |
3493 | mnt_drop_write_file(file); | |
3494 | return ret; | |
6ef5ed0d JB |
3495 | } |
3496 | ||
5af3e8cc SB |
3497 | void btrfs_get_block_group_info(struct list_head *groups_list, |
3498 | struct btrfs_ioctl_space_info *space) | |
bf5fc093 JB |
3499 | { |
3500 | struct btrfs_block_group_cache *block_group; | |
3501 | ||
3502 | space->total_bytes = 0; | |
3503 | space->used_bytes = 0; | |
3504 | space->flags = 0; | |
3505 | list_for_each_entry(block_group, groups_list, list) { | |
3506 | space->flags = block_group->flags; | |
3507 | space->total_bytes += block_group->key.offset; | |
3508 | space->used_bytes += | |
3509 | btrfs_block_group_used(&block_group->item); | |
3510 | } | |
3511 | } | |
3512 | ||
48a3b636 | 3513 | static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg) |
1406e432 JB |
3514 | { |
3515 | struct btrfs_ioctl_space_args space_args; | |
3516 | struct btrfs_ioctl_space_info space; | |
3517 | struct btrfs_ioctl_space_info *dest; | |
7fde62bf | 3518 | struct btrfs_ioctl_space_info *dest_orig; |
13f2696f | 3519 | struct btrfs_ioctl_space_info __user *user_dest; |
1406e432 | 3520 | struct btrfs_space_info *info; |
bf5fc093 JB |
3521 | u64 types[] = {BTRFS_BLOCK_GROUP_DATA, |
3522 | BTRFS_BLOCK_GROUP_SYSTEM, | |
3523 | BTRFS_BLOCK_GROUP_METADATA, | |
3524 | BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; | |
3525 | int num_types = 4; | |
7fde62bf | 3526 | int alloc_size; |
1406e432 | 3527 | int ret = 0; |
51788b1b | 3528 | u64 slot_count = 0; |
bf5fc093 | 3529 | int i, c; |
1406e432 JB |
3530 | |
3531 | if (copy_from_user(&space_args, | |
3532 | (struct btrfs_ioctl_space_args __user *)arg, | |
3533 | sizeof(space_args))) | |
3534 | return -EFAULT; | |
3535 | ||
bf5fc093 JB |
3536 | for (i = 0; i < num_types; i++) { |
3537 | struct btrfs_space_info *tmp; | |
3538 | ||
3539 | info = NULL; | |
3540 | rcu_read_lock(); | |
3541 | list_for_each_entry_rcu(tmp, &root->fs_info->space_info, | |
3542 | list) { | |
3543 | if (tmp->flags == types[i]) { | |
3544 | info = tmp; | |
3545 | break; | |
3546 | } | |
3547 | } | |
3548 | rcu_read_unlock(); | |
3549 | ||
3550 | if (!info) | |
3551 | continue; | |
3552 | ||
3553 | down_read(&info->groups_sem); | |
3554 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
3555 | if (!list_empty(&info->block_groups[c])) | |
3556 | slot_count++; | |
3557 | } | |
3558 | up_read(&info->groups_sem); | |
3559 | } | |
7fde62bf CM |
3560 | |
3561 | /* space_slots == 0 means they are asking for a count */ | |
3562 | if (space_args.space_slots == 0) { | |
3563 | space_args.total_spaces = slot_count; | |
3564 | goto out; | |
3565 | } | |
bf5fc093 | 3566 | |
51788b1b | 3567 | slot_count = min_t(u64, space_args.space_slots, slot_count); |
bf5fc093 | 3568 | |
7fde62bf | 3569 | alloc_size = sizeof(*dest) * slot_count; |
bf5fc093 | 3570 | |
7fde62bf CM |
3571 | /* we generally have at most 6 or so space infos, one for each raid |
3572 | * level. So, a whole page should be more than enough for everyone | |
3573 | */ | |
3574 | if (alloc_size > PAGE_CACHE_SIZE) | |
3575 | return -ENOMEM; | |
3576 | ||
1406e432 | 3577 | space_args.total_spaces = 0; |
7fde62bf CM |
3578 | dest = kmalloc(alloc_size, GFP_NOFS); |
3579 | if (!dest) | |
3580 | return -ENOMEM; | |
3581 | dest_orig = dest; | |
1406e432 | 3582 | |
7fde62bf | 3583 | /* now we have a buffer to copy into */ |
bf5fc093 JB |
3584 | for (i = 0; i < num_types; i++) { |
3585 | struct btrfs_space_info *tmp; | |
3586 | ||
51788b1b DR |
3587 | if (!slot_count) |
3588 | break; | |
3589 | ||
bf5fc093 JB |
3590 | info = NULL; |
3591 | rcu_read_lock(); | |
3592 | list_for_each_entry_rcu(tmp, &root->fs_info->space_info, | |
3593 | list) { | |
3594 | if (tmp->flags == types[i]) { | |
3595 | info = tmp; | |
3596 | break; | |
3597 | } | |
3598 | } | |
3599 | rcu_read_unlock(); | |
7fde62bf | 3600 | |
bf5fc093 JB |
3601 | if (!info) |
3602 | continue; | |
3603 | down_read(&info->groups_sem); | |
3604 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
3605 | if (!list_empty(&info->block_groups[c])) { | |
5af3e8cc SB |
3606 | btrfs_get_block_group_info( |
3607 | &info->block_groups[c], &space); | |
bf5fc093 JB |
3608 | memcpy(dest, &space, sizeof(space)); |
3609 | dest++; | |
3610 | space_args.total_spaces++; | |
51788b1b | 3611 | slot_count--; |
bf5fc093 | 3612 | } |
51788b1b DR |
3613 | if (!slot_count) |
3614 | break; | |
bf5fc093 JB |
3615 | } |
3616 | up_read(&info->groups_sem); | |
1406e432 | 3617 | } |
1406e432 | 3618 | |
2eec6c81 | 3619 | user_dest = (struct btrfs_ioctl_space_info __user *) |
7fde62bf CM |
3620 | (arg + sizeof(struct btrfs_ioctl_space_args)); |
3621 | ||
3622 | if (copy_to_user(user_dest, dest_orig, alloc_size)) | |
3623 | ret = -EFAULT; | |
3624 | ||
3625 | kfree(dest_orig); | |
3626 | out: | |
3627 | if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) | |
1406e432 JB |
3628 | ret = -EFAULT; |
3629 | ||
3630 | return ret; | |
3631 | } | |
3632 | ||
f46b5a66 CH |
3633 | /* |
3634 | * there are many ways the trans_start and trans_end ioctls can lead | |
3635 | * to deadlocks. They should only be used by applications that | |
3636 | * basically own the machine, and have a very in depth understanding | |
3637 | * of all the possible deadlocks and enospc problems. | |
3638 | */ | |
3639 | long btrfs_ioctl_trans_end(struct file *file) | |
3640 | { | |
496ad9aa | 3641 | struct inode *inode = file_inode(file); |
f46b5a66 CH |
3642 | struct btrfs_root *root = BTRFS_I(inode)->root; |
3643 | struct btrfs_trans_handle *trans; | |
f46b5a66 | 3644 | |
f46b5a66 | 3645 | trans = file->private_data; |
1ab86aed SW |
3646 | if (!trans) |
3647 | return -EINVAL; | |
b214107e | 3648 | file->private_data = NULL; |
9ca9ee09 | 3649 | |
1ab86aed SW |
3650 | btrfs_end_transaction(trans, root); |
3651 | ||
a4abeea4 | 3652 | atomic_dec(&root->fs_info->open_ioctl_trans); |
9ca9ee09 | 3653 | |
2a79f17e | 3654 | mnt_drop_write_file(file); |
1ab86aed | 3655 | return 0; |
f46b5a66 CH |
3656 | } |
3657 | ||
9a8c28be MX |
3658 | static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, |
3659 | void __user *argp) | |
46204592 | 3660 | { |
46204592 SW |
3661 | struct btrfs_trans_handle *trans; |
3662 | u64 transid; | |
db5b493a | 3663 | int ret; |
46204592 | 3664 | |
d4edf39b | 3665 | trans = btrfs_attach_transaction_barrier(root); |
ff7c1d33 MX |
3666 | if (IS_ERR(trans)) { |
3667 | if (PTR_ERR(trans) != -ENOENT) | |
3668 | return PTR_ERR(trans); | |
3669 | ||
3670 | /* No running transaction, don't bother */ | |
3671 | transid = root->fs_info->last_trans_committed; | |
3672 | goto out; | |
3673 | } | |
46204592 | 3674 | transid = trans->transid; |
db5b493a | 3675 | ret = btrfs_commit_transaction_async(trans, root, 0); |
8b2b2d3c TI |
3676 | if (ret) { |
3677 | btrfs_end_transaction(trans, root); | |
db5b493a | 3678 | return ret; |
8b2b2d3c | 3679 | } |
ff7c1d33 | 3680 | out: |
46204592 SW |
3681 | if (argp) |
3682 | if (copy_to_user(argp, &transid, sizeof(transid))) | |
3683 | return -EFAULT; | |
3684 | return 0; | |
3685 | } | |
3686 | ||
9a8c28be MX |
3687 | static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root, |
3688 | void __user *argp) | |
46204592 | 3689 | { |
46204592 SW |
3690 | u64 transid; |
3691 | ||
3692 | if (argp) { | |
3693 | if (copy_from_user(&transid, argp, sizeof(transid))) | |
3694 | return -EFAULT; | |
3695 | } else { | |
3696 | transid = 0; /* current trans */ | |
3697 | } | |
3698 | return btrfs_wait_for_commit(root, transid); | |
3699 | } | |
3700 | ||
b8e95489 | 3701 | static long btrfs_ioctl_scrub(struct file *file, void __user *arg) |
475f6387 | 3702 | { |
496ad9aa | 3703 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
475f6387 | 3704 | struct btrfs_ioctl_scrub_args *sa; |
b8e95489 | 3705 | int ret; |
475f6387 JS |
3706 | |
3707 | if (!capable(CAP_SYS_ADMIN)) | |
3708 | return -EPERM; | |
3709 | ||
3710 | sa = memdup_user(arg, sizeof(*sa)); | |
3711 | if (IS_ERR(sa)) | |
3712 | return PTR_ERR(sa); | |
3713 | ||
b8e95489 MX |
3714 | if (!(sa->flags & BTRFS_SCRUB_READONLY)) { |
3715 | ret = mnt_want_write_file(file); | |
3716 | if (ret) | |
3717 | goto out; | |
3718 | } | |
3719 | ||
aa1b8cd4 | 3720 | ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end, |
63a212ab SB |
3721 | &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, |
3722 | 0); | |
475f6387 JS |
3723 | |
3724 | if (copy_to_user(arg, sa, sizeof(*sa))) | |
3725 | ret = -EFAULT; | |
3726 | ||
b8e95489 MX |
3727 | if (!(sa->flags & BTRFS_SCRUB_READONLY)) |
3728 | mnt_drop_write_file(file); | |
3729 | out: | |
475f6387 JS |
3730 | kfree(sa); |
3731 | return ret; | |
3732 | } | |
3733 | ||
3734 | static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg) | |
3735 | { | |
3736 | if (!capable(CAP_SYS_ADMIN)) | |
3737 | return -EPERM; | |
3738 | ||
aa1b8cd4 | 3739 | return btrfs_scrub_cancel(root->fs_info); |
475f6387 JS |
3740 | } |
3741 | ||
3742 | static long btrfs_ioctl_scrub_progress(struct btrfs_root *root, | |
3743 | void __user *arg) | |
3744 | { | |
3745 | struct btrfs_ioctl_scrub_args *sa; | |
3746 | int ret; | |
3747 | ||
3748 | if (!capable(CAP_SYS_ADMIN)) | |
3749 | return -EPERM; | |
3750 | ||
3751 | sa = memdup_user(arg, sizeof(*sa)); | |
3752 | if (IS_ERR(sa)) | |
3753 | return PTR_ERR(sa); | |
3754 | ||
3755 | ret = btrfs_scrub_progress(root, sa->devid, &sa->progress); | |
3756 | ||
3757 | if (copy_to_user(arg, sa, sizeof(*sa))) | |
3758 | ret = -EFAULT; | |
3759 | ||
3760 | kfree(sa); | |
3761 | return ret; | |
3762 | } | |
3763 | ||
c11d2c23 | 3764 | static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root, |
b27f7c0c | 3765 | void __user *arg) |
c11d2c23 SB |
3766 | { |
3767 | struct btrfs_ioctl_get_dev_stats *sa; | |
3768 | int ret; | |
3769 | ||
c11d2c23 SB |
3770 | sa = memdup_user(arg, sizeof(*sa)); |
3771 | if (IS_ERR(sa)) | |
3772 | return PTR_ERR(sa); | |
3773 | ||
b27f7c0c DS |
3774 | if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { |
3775 | kfree(sa); | |
3776 | return -EPERM; | |
3777 | } | |
3778 | ||
3779 | ret = btrfs_get_dev_stats(root, sa); | |
c11d2c23 SB |
3780 | |
3781 | if (copy_to_user(arg, sa, sizeof(*sa))) | |
3782 | ret = -EFAULT; | |
3783 | ||
3784 | kfree(sa); | |
3785 | return ret; | |
3786 | } | |
3787 | ||
3f6bcfbd SB |
3788 | static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg) |
3789 | { | |
3790 | struct btrfs_ioctl_dev_replace_args *p; | |
3791 | int ret; | |
3792 | ||
3793 | if (!capable(CAP_SYS_ADMIN)) | |
3794 | return -EPERM; | |
3795 | ||
3796 | p = memdup_user(arg, sizeof(*p)); | |
3797 | if (IS_ERR(p)) | |
3798 | return PTR_ERR(p); | |
3799 | ||
3800 | switch (p->cmd) { | |
3801 | case BTRFS_IOCTL_DEV_REPLACE_CMD_START: | |
adfa97cb ID |
3802 | if (root->fs_info->sb->s_flags & MS_RDONLY) { |
3803 | ret = -EROFS; | |
3804 | goto out; | |
3805 | } | |
3f6bcfbd SB |
3806 | if (atomic_xchg( |
3807 | &root->fs_info->mutually_exclusive_operation_running, | |
3808 | 1)) { | |
e57138b3 | 3809 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
3f6bcfbd SB |
3810 | } else { |
3811 | ret = btrfs_dev_replace_start(root, p); | |
3812 | atomic_set( | |
3813 | &root->fs_info->mutually_exclusive_operation_running, | |
3814 | 0); | |
3815 | } | |
3816 | break; | |
3817 | case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: | |
3818 | btrfs_dev_replace_status(root->fs_info, p); | |
3819 | ret = 0; | |
3820 | break; | |
3821 | case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: | |
3822 | ret = btrfs_dev_replace_cancel(root->fs_info, p); | |
3823 | break; | |
3824 | default: | |
3825 | ret = -EINVAL; | |
3826 | break; | |
3827 | } | |
3828 | ||
3829 | if (copy_to_user(arg, p, sizeof(*p))) | |
3830 | ret = -EFAULT; | |
adfa97cb | 3831 | out: |
3f6bcfbd SB |
3832 | kfree(p); |
3833 | return ret; | |
3834 | } | |
3835 | ||
d7728c96 JS |
3836 | static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) |
3837 | { | |
3838 | int ret = 0; | |
3839 | int i; | |
740c3d22 | 3840 | u64 rel_ptr; |
d7728c96 | 3841 | int size; |
806468f8 | 3842 | struct btrfs_ioctl_ino_path_args *ipa = NULL; |
d7728c96 JS |
3843 | struct inode_fs_paths *ipath = NULL; |
3844 | struct btrfs_path *path; | |
3845 | ||
82b22ac8 | 3846 | if (!capable(CAP_DAC_READ_SEARCH)) |
d7728c96 JS |
3847 | return -EPERM; |
3848 | ||
3849 | path = btrfs_alloc_path(); | |
3850 | if (!path) { | |
3851 | ret = -ENOMEM; | |
3852 | goto out; | |
3853 | } | |
3854 | ||
3855 | ipa = memdup_user(arg, sizeof(*ipa)); | |
3856 | if (IS_ERR(ipa)) { | |
3857 | ret = PTR_ERR(ipa); | |
3858 | ipa = NULL; | |
3859 | goto out; | |
3860 | } | |
3861 | ||
3862 | size = min_t(u32, ipa->size, 4096); | |
3863 | ipath = init_ipath(size, root, path); | |
3864 | if (IS_ERR(ipath)) { | |
3865 | ret = PTR_ERR(ipath); | |
3866 | ipath = NULL; | |
3867 | goto out; | |
3868 | } | |
3869 | ||
3870 | ret = paths_from_inode(ipa->inum, ipath); | |
3871 | if (ret < 0) | |
3872 | goto out; | |
3873 | ||
3874 | for (i = 0; i < ipath->fspath->elem_cnt; ++i) { | |
745c4d8e JM |
3875 | rel_ptr = ipath->fspath->val[i] - |
3876 | (u64)(unsigned long)ipath->fspath->val; | |
740c3d22 | 3877 | ipath->fspath->val[i] = rel_ptr; |
d7728c96 JS |
3878 | } |
3879 | ||
745c4d8e JM |
3880 | ret = copy_to_user((void *)(unsigned long)ipa->fspath, |
3881 | (void *)(unsigned long)ipath->fspath, size); | |
d7728c96 JS |
3882 | if (ret) { |
3883 | ret = -EFAULT; | |
3884 | goto out; | |
3885 | } | |
3886 | ||
3887 | out: | |
3888 | btrfs_free_path(path); | |
3889 | free_ipath(ipath); | |
3890 | kfree(ipa); | |
3891 | ||
3892 | return ret; | |
3893 | } | |
3894 | ||
3895 | static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) | |
3896 | { | |
3897 | struct btrfs_data_container *inodes = ctx; | |
3898 | const size_t c = 3 * sizeof(u64); | |
3899 | ||
3900 | if (inodes->bytes_left >= c) { | |
3901 | inodes->bytes_left -= c; | |
3902 | inodes->val[inodes->elem_cnt] = inum; | |
3903 | inodes->val[inodes->elem_cnt + 1] = offset; | |
3904 | inodes->val[inodes->elem_cnt + 2] = root; | |
3905 | inodes->elem_cnt += 3; | |
3906 | } else { | |
3907 | inodes->bytes_missing += c - inodes->bytes_left; | |
3908 | inodes->bytes_left = 0; | |
3909 | inodes->elem_missed += 3; | |
3910 | } | |
3911 | ||
3912 | return 0; | |
3913 | } | |
3914 | ||
3915 | static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root, | |
3916 | void __user *arg) | |
3917 | { | |
3918 | int ret = 0; | |
3919 | int size; | |
d7728c96 JS |
3920 | struct btrfs_ioctl_logical_ino_args *loi; |
3921 | struct btrfs_data_container *inodes = NULL; | |
3922 | struct btrfs_path *path = NULL; | |
d7728c96 JS |
3923 | |
3924 | if (!capable(CAP_SYS_ADMIN)) | |
3925 | return -EPERM; | |
3926 | ||
3927 | loi = memdup_user(arg, sizeof(*loi)); | |
3928 | if (IS_ERR(loi)) { | |
3929 | ret = PTR_ERR(loi); | |
3930 | loi = NULL; | |
3931 | goto out; | |
3932 | } | |
3933 | ||
3934 | path = btrfs_alloc_path(); | |
3935 | if (!path) { | |
3936 | ret = -ENOMEM; | |
3937 | goto out; | |
3938 | } | |
3939 | ||
425d17a2 | 3940 | size = min_t(u32, loi->size, 64 * 1024); |
d7728c96 JS |
3941 | inodes = init_data_container(size); |
3942 | if (IS_ERR(inodes)) { | |
3943 | ret = PTR_ERR(inodes); | |
3944 | inodes = NULL; | |
3945 | goto out; | |
3946 | } | |
3947 | ||
df031f07 LB |
3948 | ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path, |
3949 | build_ino_list, inodes); | |
3950 | if (ret == -EINVAL) | |
d7728c96 JS |
3951 | ret = -ENOENT; |
3952 | if (ret < 0) | |
3953 | goto out; | |
3954 | ||
745c4d8e JM |
3955 | ret = copy_to_user((void *)(unsigned long)loi->inodes, |
3956 | (void *)(unsigned long)inodes, size); | |
d7728c96 JS |
3957 | if (ret) |
3958 | ret = -EFAULT; | |
3959 | ||
3960 | out: | |
3961 | btrfs_free_path(path); | |
425d17a2 | 3962 | vfree(inodes); |
d7728c96 JS |
3963 | kfree(loi); |
3964 | ||
3965 | return ret; | |
3966 | } | |
3967 | ||
19a39dce | 3968 | void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock, |
c9e9f97b ID |
3969 | struct btrfs_ioctl_balance_args *bargs) |
3970 | { | |
3971 | struct btrfs_balance_control *bctl = fs_info->balance_ctl; | |
3972 | ||
3973 | bargs->flags = bctl->flags; | |
3974 | ||
837d5b6e ID |
3975 | if (atomic_read(&fs_info->balance_running)) |
3976 | bargs->state |= BTRFS_BALANCE_STATE_RUNNING; | |
3977 | if (atomic_read(&fs_info->balance_pause_req)) | |
3978 | bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; | |
a7e99c69 ID |
3979 | if (atomic_read(&fs_info->balance_cancel_req)) |
3980 | bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; | |
837d5b6e | 3981 | |
c9e9f97b ID |
3982 | memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); |
3983 | memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); | |
3984 | memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); | |
19a39dce ID |
3985 | |
3986 | if (lock) { | |
3987 | spin_lock(&fs_info->balance_lock); | |
3988 | memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); | |
3989 | spin_unlock(&fs_info->balance_lock); | |
3990 | } else { | |
3991 | memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); | |
3992 | } | |
c9e9f97b ID |
3993 | } |
3994 | ||
9ba1f6e4 | 3995 | static long btrfs_ioctl_balance(struct file *file, void __user *arg) |
c9e9f97b | 3996 | { |
496ad9aa | 3997 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
c9e9f97b ID |
3998 | struct btrfs_fs_info *fs_info = root->fs_info; |
3999 | struct btrfs_ioctl_balance_args *bargs; | |
4000 | struct btrfs_balance_control *bctl; | |
ed0fb78f | 4001 | bool need_unlock; /* for mut. excl. ops lock */ |
c9e9f97b ID |
4002 | int ret; |
4003 | ||
4004 | if (!capable(CAP_SYS_ADMIN)) | |
4005 | return -EPERM; | |
4006 | ||
e54bfa31 | 4007 | ret = mnt_want_write_file(file); |
9ba1f6e4 LB |
4008 | if (ret) |
4009 | return ret; | |
4010 | ||
ed0fb78f ID |
4011 | again: |
4012 | if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { | |
4013 | mutex_lock(&fs_info->volume_mutex); | |
4014 | mutex_lock(&fs_info->balance_mutex); | |
4015 | need_unlock = true; | |
4016 | goto locked; | |
4017 | } | |
4018 | ||
4019 | /* | |
4020 | * mut. excl. ops lock is locked. Three possibilites: | |
4021 | * (1) some other op is running | |
4022 | * (2) balance is running | |
4023 | * (3) balance is paused -- special case (think resume) | |
4024 | */ | |
c9e9f97b | 4025 | mutex_lock(&fs_info->balance_mutex); |
ed0fb78f ID |
4026 | if (fs_info->balance_ctl) { |
4027 | /* this is either (2) or (3) */ | |
4028 | if (!atomic_read(&fs_info->balance_running)) { | |
4029 | mutex_unlock(&fs_info->balance_mutex); | |
4030 | if (!mutex_trylock(&fs_info->volume_mutex)) | |
4031 | goto again; | |
4032 | mutex_lock(&fs_info->balance_mutex); | |
4033 | ||
4034 | if (fs_info->balance_ctl && | |
4035 | !atomic_read(&fs_info->balance_running)) { | |
4036 | /* this is (3) */ | |
4037 | need_unlock = false; | |
4038 | goto locked; | |
4039 | } | |
4040 | ||
4041 | mutex_unlock(&fs_info->balance_mutex); | |
4042 | mutex_unlock(&fs_info->volume_mutex); | |
4043 | goto again; | |
4044 | } else { | |
4045 | /* this is (2) */ | |
4046 | mutex_unlock(&fs_info->balance_mutex); | |
4047 | ret = -EINPROGRESS; | |
4048 | goto out; | |
4049 | } | |
4050 | } else { | |
4051 | /* this is (1) */ | |
4052 | mutex_unlock(&fs_info->balance_mutex); | |
e57138b3 | 4053 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
ed0fb78f ID |
4054 | goto out; |
4055 | } | |
4056 | ||
4057 | locked: | |
4058 | BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running)); | |
c9e9f97b ID |
4059 | |
4060 | if (arg) { | |
4061 | bargs = memdup_user(arg, sizeof(*bargs)); | |
4062 | if (IS_ERR(bargs)) { | |
4063 | ret = PTR_ERR(bargs); | |
ed0fb78f | 4064 | goto out_unlock; |
c9e9f97b | 4065 | } |
de322263 ID |
4066 | |
4067 | if (bargs->flags & BTRFS_BALANCE_RESUME) { | |
4068 | if (!fs_info->balance_ctl) { | |
4069 | ret = -ENOTCONN; | |
4070 | goto out_bargs; | |
4071 | } | |
4072 | ||
4073 | bctl = fs_info->balance_ctl; | |
4074 | spin_lock(&fs_info->balance_lock); | |
4075 | bctl->flags |= BTRFS_BALANCE_RESUME; | |
4076 | spin_unlock(&fs_info->balance_lock); | |
4077 | ||
4078 | goto do_balance; | |
4079 | } | |
c9e9f97b ID |
4080 | } else { |
4081 | bargs = NULL; | |
4082 | } | |
4083 | ||
ed0fb78f | 4084 | if (fs_info->balance_ctl) { |
837d5b6e ID |
4085 | ret = -EINPROGRESS; |
4086 | goto out_bargs; | |
4087 | } | |
4088 | ||
c9e9f97b ID |
4089 | bctl = kzalloc(sizeof(*bctl), GFP_NOFS); |
4090 | if (!bctl) { | |
4091 | ret = -ENOMEM; | |
4092 | goto out_bargs; | |
4093 | } | |
4094 | ||
4095 | bctl->fs_info = fs_info; | |
4096 | if (arg) { | |
4097 | memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); | |
4098 | memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); | |
4099 | memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); | |
4100 | ||
4101 | bctl->flags = bargs->flags; | |
f43ffb60 ID |
4102 | } else { |
4103 | /* balance everything - no filters */ | |
4104 | bctl->flags |= BTRFS_BALANCE_TYPE_MASK; | |
c9e9f97b ID |
4105 | } |
4106 | ||
de322263 | 4107 | do_balance: |
c9e9f97b | 4108 | /* |
ed0fb78f ID |
4109 | * Ownership of bctl and mutually_exclusive_operation_running |
4110 | * goes to to btrfs_balance. bctl is freed in __cancel_balance, | |
4111 | * or, if restriper was paused all the way until unmount, in | |
4112 | * free_fs_info. mutually_exclusive_operation_running is | |
4113 | * cleared in __cancel_balance. | |
c9e9f97b | 4114 | */ |
ed0fb78f ID |
4115 | need_unlock = false; |
4116 | ||
4117 | ret = btrfs_balance(bctl, bargs); | |
4118 | ||
c9e9f97b ID |
4119 | if (arg) { |
4120 | if (copy_to_user(arg, bargs, sizeof(*bargs))) | |
4121 | ret = -EFAULT; | |
4122 | } | |
4123 | ||
4124 | out_bargs: | |
4125 | kfree(bargs); | |
ed0fb78f | 4126 | out_unlock: |
c9e9f97b ID |
4127 | mutex_unlock(&fs_info->balance_mutex); |
4128 | mutex_unlock(&fs_info->volume_mutex); | |
ed0fb78f ID |
4129 | if (need_unlock) |
4130 | atomic_set(&fs_info->mutually_exclusive_operation_running, 0); | |
4131 | out: | |
e54bfa31 | 4132 | mnt_drop_write_file(file); |
c9e9f97b ID |
4133 | return ret; |
4134 | } | |
4135 | ||
837d5b6e ID |
4136 | static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd) |
4137 | { | |
4138 | if (!capable(CAP_SYS_ADMIN)) | |
4139 | return -EPERM; | |
4140 | ||
4141 | switch (cmd) { | |
4142 | case BTRFS_BALANCE_CTL_PAUSE: | |
4143 | return btrfs_pause_balance(root->fs_info); | |
a7e99c69 ID |
4144 | case BTRFS_BALANCE_CTL_CANCEL: |
4145 | return btrfs_cancel_balance(root->fs_info); | |
837d5b6e ID |
4146 | } |
4147 | ||
4148 | return -EINVAL; | |
4149 | } | |
4150 | ||
19a39dce ID |
4151 | static long btrfs_ioctl_balance_progress(struct btrfs_root *root, |
4152 | void __user *arg) | |
4153 | { | |
4154 | struct btrfs_fs_info *fs_info = root->fs_info; | |
4155 | struct btrfs_ioctl_balance_args *bargs; | |
4156 | int ret = 0; | |
4157 | ||
4158 | if (!capable(CAP_SYS_ADMIN)) | |
4159 | return -EPERM; | |
4160 | ||
4161 | mutex_lock(&fs_info->balance_mutex); | |
4162 | if (!fs_info->balance_ctl) { | |
4163 | ret = -ENOTCONN; | |
4164 | goto out; | |
4165 | } | |
4166 | ||
4167 | bargs = kzalloc(sizeof(*bargs), GFP_NOFS); | |
4168 | if (!bargs) { | |
4169 | ret = -ENOMEM; | |
4170 | goto out; | |
4171 | } | |
4172 | ||
4173 | update_ioctl_balance_args(fs_info, 1, bargs); | |
4174 | ||
4175 | if (copy_to_user(arg, bargs, sizeof(*bargs))) | |
4176 | ret = -EFAULT; | |
4177 | ||
4178 | kfree(bargs); | |
4179 | out: | |
4180 | mutex_unlock(&fs_info->balance_mutex); | |
4181 | return ret; | |
4182 | } | |
4183 | ||
905b0dda | 4184 | static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) |
5d13a37b | 4185 | { |
496ad9aa | 4186 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
5d13a37b AJ |
4187 | struct btrfs_ioctl_quota_ctl_args *sa; |
4188 | struct btrfs_trans_handle *trans = NULL; | |
4189 | int ret; | |
4190 | int err; | |
4191 | ||
4192 | if (!capable(CAP_SYS_ADMIN)) | |
4193 | return -EPERM; | |
4194 | ||
905b0dda MX |
4195 | ret = mnt_want_write_file(file); |
4196 | if (ret) | |
4197 | return ret; | |
5d13a37b AJ |
4198 | |
4199 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4200 | if (IS_ERR(sa)) { |
4201 | ret = PTR_ERR(sa); | |
4202 | goto drop_write; | |
4203 | } | |
5d13a37b | 4204 | |
7708f029 | 4205 | down_write(&root->fs_info->subvol_sem); |
2f232036 JS |
4206 | trans = btrfs_start_transaction(root->fs_info->tree_root, 2); |
4207 | if (IS_ERR(trans)) { | |
4208 | ret = PTR_ERR(trans); | |
4209 | goto out; | |
5d13a37b AJ |
4210 | } |
4211 | ||
4212 | switch (sa->cmd) { | |
4213 | case BTRFS_QUOTA_CTL_ENABLE: | |
4214 | ret = btrfs_quota_enable(trans, root->fs_info); | |
4215 | break; | |
4216 | case BTRFS_QUOTA_CTL_DISABLE: | |
4217 | ret = btrfs_quota_disable(trans, root->fs_info); | |
4218 | break; | |
5d13a37b AJ |
4219 | default: |
4220 | ret = -EINVAL; | |
4221 | break; | |
4222 | } | |
4223 | ||
2f232036 JS |
4224 | err = btrfs_commit_transaction(trans, root->fs_info->tree_root); |
4225 | if (err && !ret) | |
4226 | ret = err; | |
5d13a37b AJ |
4227 | out: |
4228 | kfree(sa); | |
7708f029 | 4229 | up_write(&root->fs_info->subvol_sem); |
905b0dda MX |
4230 | drop_write: |
4231 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4232 | return ret; |
4233 | } | |
4234 | ||
905b0dda | 4235 | static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) |
5d13a37b | 4236 | { |
496ad9aa | 4237 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
5d13a37b AJ |
4238 | struct btrfs_ioctl_qgroup_assign_args *sa; |
4239 | struct btrfs_trans_handle *trans; | |
4240 | int ret; | |
4241 | int err; | |
4242 | ||
4243 | if (!capable(CAP_SYS_ADMIN)) | |
4244 | return -EPERM; | |
4245 | ||
905b0dda MX |
4246 | ret = mnt_want_write_file(file); |
4247 | if (ret) | |
4248 | return ret; | |
5d13a37b AJ |
4249 | |
4250 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4251 | if (IS_ERR(sa)) { |
4252 | ret = PTR_ERR(sa); | |
4253 | goto drop_write; | |
4254 | } | |
5d13a37b AJ |
4255 | |
4256 | trans = btrfs_join_transaction(root); | |
4257 | if (IS_ERR(trans)) { | |
4258 | ret = PTR_ERR(trans); | |
4259 | goto out; | |
4260 | } | |
4261 | ||
4262 | /* FIXME: check if the IDs really exist */ | |
4263 | if (sa->assign) { | |
4264 | ret = btrfs_add_qgroup_relation(trans, root->fs_info, | |
4265 | sa->src, sa->dst); | |
4266 | } else { | |
4267 | ret = btrfs_del_qgroup_relation(trans, root->fs_info, | |
4268 | sa->src, sa->dst); | |
4269 | } | |
4270 | ||
4271 | err = btrfs_end_transaction(trans, root); | |
4272 | if (err && !ret) | |
4273 | ret = err; | |
4274 | ||
4275 | out: | |
4276 | kfree(sa); | |
905b0dda MX |
4277 | drop_write: |
4278 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4279 | return ret; |
4280 | } | |
4281 | ||
905b0dda | 4282 | static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) |
5d13a37b | 4283 | { |
496ad9aa | 4284 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
5d13a37b AJ |
4285 | struct btrfs_ioctl_qgroup_create_args *sa; |
4286 | struct btrfs_trans_handle *trans; | |
4287 | int ret; | |
4288 | int err; | |
4289 | ||
4290 | if (!capable(CAP_SYS_ADMIN)) | |
4291 | return -EPERM; | |
4292 | ||
905b0dda MX |
4293 | ret = mnt_want_write_file(file); |
4294 | if (ret) | |
4295 | return ret; | |
5d13a37b AJ |
4296 | |
4297 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4298 | if (IS_ERR(sa)) { |
4299 | ret = PTR_ERR(sa); | |
4300 | goto drop_write; | |
4301 | } | |
5d13a37b | 4302 | |
d86e56cf MX |
4303 | if (!sa->qgroupid) { |
4304 | ret = -EINVAL; | |
4305 | goto out; | |
4306 | } | |
4307 | ||
5d13a37b AJ |
4308 | trans = btrfs_join_transaction(root); |
4309 | if (IS_ERR(trans)) { | |
4310 | ret = PTR_ERR(trans); | |
4311 | goto out; | |
4312 | } | |
4313 | ||
4314 | /* FIXME: check if the IDs really exist */ | |
4315 | if (sa->create) { | |
4316 | ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid, | |
4317 | NULL); | |
4318 | } else { | |
4319 | ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid); | |
4320 | } | |
4321 | ||
4322 | err = btrfs_end_transaction(trans, root); | |
4323 | if (err && !ret) | |
4324 | ret = err; | |
4325 | ||
4326 | out: | |
4327 | kfree(sa); | |
905b0dda MX |
4328 | drop_write: |
4329 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4330 | return ret; |
4331 | } | |
4332 | ||
905b0dda | 4333 | static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) |
5d13a37b | 4334 | { |
496ad9aa | 4335 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
5d13a37b AJ |
4336 | struct btrfs_ioctl_qgroup_limit_args *sa; |
4337 | struct btrfs_trans_handle *trans; | |
4338 | int ret; | |
4339 | int err; | |
4340 | u64 qgroupid; | |
4341 | ||
4342 | if (!capable(CAP_SYS_ADMIN)) | |
4343 | return -EPERM; | |
4344 | ||
905b0dda MX |
4345 | ret = mnt_want_write_file(file); |
4346 | if (ret) | |
4347 | return ret; | |
5d13a37b AJ |
4348 | |
4349 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4350 | if (IS_ERR(sa)) { |
4351 | ret = PTR_ERR(sa); | |
4352 | goto drop_write; | |
4353 | } | |
5d13a37b AJ |
4354 | |
4355 | trans = btrfs_join_transaction(root); | |
4356 | if (IS_ERR(trans)) { | |
4357 | ret = PTR_ERR(trans); | |
4358 | goto out; | |
4359 | } | |
4360 | ||
4361 | qgroupid = sa->qgroupid; | |
4362 | if (!qgroupid) { | |
4363 | /* take the current subvol as qgroup */ | |
4364 | qgroupid = root->root_key.objectid; | |
4365 | } | |
4366 | ||
4367 | /* FIXME: check if the IDs really exist */ | |
4368 | ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim); | |
4369 | ||
4370 | err = btrfs_end_transaction(trans, root); | |
4371 | if (err && !ret) | |
4372 | ret = err; | |
4373 | ||
4374 | out: | |
4375 | kfree(sa); | |
905b0dda MX |
4376 | drop_write: |
4377 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4378 | return ret; |
4379 | } | |
4380 | ||
2f232036 JS |
4381 | static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) |
4382 | { | |
6d0379ec | 4383 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
2f232036 JS |
4384 | struct btrfs_ioctl_quota_rescan_args *qsa; |
4385 | int ret; | |
4386 | ||
4387 | if (!capable(CAP_SYS_ADMIN)) | |
4388 | return -EPERM; | |
4389 | ||
4390 | ret = mnt_want_write_file(file); | |
4391 | if (ret) | |
4392 | return ret; | |
4393 | ||
4394 | qsa = memdup_user(arg, sizeof(*qsa)); | |
4395 | if (IS_ERR(qsa)) { | |
4396 | ret = PTR_ERR(qsa); | |
4397 | goto drop_write; | |
4398 | } | |
4399 | ||
4400 | if (qsa->flags) { | |
4401 | ret = -EINVAL; | |
4402 | goto out; | |
4403 | } | |
4404 | ||
4405 | ret = btrfs_qgroup_rescan(root->fs_info); | |
4406 | ||
4407 | out: | |
4408 | kfree(qsa); | |
4409 | drop_write: | |
4410 | mnt_drop_write_file(file); | |
4411 | return ret; | |
4412 | } | |
4413 | ||
4414 | static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) | |
4415 | { | |
6d0379ec | 4416 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
2f232036 JS |
4417 | struct btrfs_ioctl_quota_rescan_args *qsa; |
4418 | int ret = 0; | |
4419 | ||
4420 | if (!capable(CAP_SYS_ADMIN)) | |
4421 | return -EPERM; | |
4422 | ||
4423 | qsa = kzalloc(sizeof(*qsa), GFP_NOFS); | |
4424 | if (!qsa) | |
4425 | return -ENOMEM; | |
4426 | ||
4427 | if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { | |
4428 | qsa->flags = 1; | |
4429 | qsa->progress = root->fs_info->qgroup_rescan_progress.objectid; | |
4430 | } | |
4431 | ||
4432 | if (copy_to_user(arg, qsa, sizeof(*qsa))) | |
4433 | ret = -EFAULT; | |
4434 | ||
4435 | kfree(qsa); | |
4436 | return ret; | |
4437 | } | |
4438 | ||
57254b6e JS |
4439 | static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) |
4440 | { | |
54563d41 | 4441 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
57254b6e JS |
4442 | |
4443 | if (!capable(CAP_SYS_ADMIN)) | |
4444 | return -EPERM; | |
4445 | ||
4446 | return btrfs_qgroup_wait_for_completion(root->fs_info); | |
4447 | } | |
4448 | ||
abccd00f HM |
4449 | static long _btrfs_ioctl_set_received_subvol(struct file *file, |
4450 | struct btrfs_ioctl_received_subvol_args *sa) | |
8ea05e3a | 4451 | { |
496ad9aa | 4452 | struct inode *inode = file_inode(file); |
8ea05e3a AB |
4453 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4454 | struct btrfs_root_item *root_item = &root->root_item; | |
4455 | struct btrfs_trans_handle *trans; | |
4456 | struct timespec ct = CURRENT_TIME; | |
4457 | int ret = 0; | |
dd5f9615 | 4458 | int received_uuid_changed; |
8ea05e3a | 4459 | |
bd60ea0f DS |
4460 | if (!inode_owner_or_capable(inode)) |
4461 | return -EPERM; | |
4462 | ||
8ea05e3a AB |
4463 | ret = mnt_want_write_file(file); |
4464 | if (ret < 0) | |
4465 | return ret; | |
4466 | ||
4467 | down_write(&root->fs_info->subvol_sem); | |
4468 | ||
4469 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { | |
4470 | ret = -EINVAL; | |
4471 | goto out; | |
4472 | } | |
4473 | ||
4474 | if (btrfs_root_readonly(root)) { | |
4475 | ret = -EROFS; | |
4476 | goto out; | |
4477 | } | |
4478 | ||
dd5f9615 SB |
4479 | /* |
4480 | * 1 - root item | |
4481 | * 2 - uuid items (received uuid + subvol uuid) | |
4482 | */ | |
4483 | trans = btrfs_start_transaction(root, 3); | |
8ea05e3a AB |
4484 | if (IS_ERR(trans)) { |
4485 | ret = PTR_ERR(trans); | |
4486 | trans = NULL; | |
4487 | goto out; | |
4488 | } | |
4489 | ||
4490 | sa->rtransid = trans->transid; | |
4491 | sa->rtime.sec = ct.tv_sec; | |
4492 | sa->rtime.nsec = ct.tv_nsec; | |
4493 | ||
dd5f9615 SB |
4494 | received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, |
4495 | BTRFS_UUID_SIZE); | |
4496 | if (received_uuid_changed && | |
4497 | !btrfs_is_empty_uuid(root_item->received_uuid)) | |
4498 | btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, | |
4499 | root_item->received_uuid, | |
4500 | BTRFS_UUID_KEY_RECEIVED_SUBVOL, | |
4501 | root->root_key.objectid); | |
8ea05e3a AB |
4502 | memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); |
4503 | btrfs_set_root_stransid(root_item, sa->stransid); | |
4504 | btrfs_set_root_rtransid(root_item, sa->rtransid); | |
3cae210f QW |
4505 | btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); |
4506 | btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); | |
4507 | btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); | |
4508 | btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); | |
8ea05e3a AB |
4509 | |
4510 | ret = btrfs_update_root(trans, root->fs_info->tree_root, | |
4511 | &root->root_key, &root->root_item); | |
4512 | if (ret < 0) { | |
4513 | btrfs_end_transaction(trans, root); | |
8ea05e3a | 4514 | goto out; |
dd5f9615 SB |
4515 | } |
4516 | if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { | |
4517 | ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root, | |
4518 | sa->uuid, | |
4519 | BTRFS_UUID_KEY_RECEIVED_SUBVOL, | |
4520 | root->root_key.objectid); | |
4521 | if (ret < 0 && ret != -EEXIST) { | |
4522 | btrfs_abort_transaction(trans, root, ret); | |
8ea05e3a | 4523 | goto out; |
dd5f9615 SB |
4524 | } |
4525 | } | |
4526 | ret = btrfs_commit_transaction(trans, root); | |
4527 | if (ret < 0) { | |
4528 | btrfs_abort_transaction(trans, root, ret); | |
4529 | goto out; | |
8ea05e3a AB |
4530 | } |
4531 | ||
abccd00f HM |
4532 | out: |
4533 | up_write(&root->fs_info->subvol_sem); | |
4534 | mnt_drop_write_file(file); | |
4535 | return ret; | |
4536 | } | |
4537 | ||
4538 | #ifdef CONFIG_64BIT | |
4539 | static long btrfs_ioctl_set_received_subvol_32(struct file *file, | |
4540 | void __user *arg) | |
4541 | { | |
4542 | struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; | |
4543 | struct btrfs_ioctl_received_subvol_args *args64 = NULL; | |
4544 | int ret = 0; | |
4545 | ||
4546 | args32 = memdup_user(arg, sizeof(*args32)); | |
4547 | if (IS_ERR(args32)) { | |
4548 | ret = PTR_ERR(args32); | |
4549 | args32 = NULL; | |
4550 | goto out; | |
4551 | } | |
4552 | ||
4553 | args64 = kmalloc(sizeof(*args64), GFP_NOFS); | |
4554 | if (IS_ERR(args64)) { | |
4555 | ret = PTR_ERR(args64); | |
4556 | args64 = NULL; | |
4557 | goto out; | |
4558 | } | |
4559 | ||
4560 | memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); | |
4561 | args64->stransid = args32->stransid; | |
4562 | args64->rtransid = args32->rtransid; | |
4563 | args64->stime.sec = args32->stime.sec; | |
4564 | args64->stime.nsec = args32->stime.nsec; | |
4565 | args64->rtime.sec = args32->rtime.sec; | |
4566 | args64->rtime.nsec = args32->rtime.nsec; | |
4567 | args64->flags = args32->flags; | |
4568 | ||
4569 | ret = _btrfs_ioctl_set_received_subvol(file, args64); | |
4570 | if (ret) | |
4571 | goto out; | |
4572 | ||
4573 | memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); | |
4574 | args32->stransid = args64->stransid; | |
4575 | args32->rtransid = args64->rtransid; | |
4576 | args32->stime.sec = args64->stime.sec; | |
4577 | args32->stime.nsec = args64->stime.nsec; | |
4578 | args32->rtime.sec = args64->rtime.sec; | |
4579 | args32->rtime.nsec = args64->rtime.nsec; | |
4580 | args32->flags = args64->flags; | |
4581 | ||
4582 | ret = copy_to_user(arg, args32, sizeof(*args32)); | |
4583 | if (ret) | |
4584 | ret = -EFAULT; | |
4585 | ||
4586 | out: | |
4587 | kfree(args32); | |
4588 | kfree(args64); | |
4589 | return ret; | |
4590 | } | |
4591 | #endif | |
4592 | ||
4593 | static long btrfs_ioctl_set_received_subvol(struct file *file, | |
4594 | void __user *arg) | |
4595 | { | |
4596 | struct btrfs_ioctl_received_subvol_args *sa = NULL; | |
4597 | int ret = 0; | |
4598 | ||
4599 | sa = memdup_user(arg, sizeof(*sa)); | |
4600 | if (IS_ERR(sa)) { | |
4601 | ret = PTR_ERR(sa); | |
4602 | sa = NULL; | |
4603 | goto out; | |
4604 | } | |
4605 | ||
4606 | ret = _btrfs_ioctl_set_received_subvol(file, sa); | |
4607 | ||
4608 | if (ret) | |
4609 | goto out; | |
4610 | ||
8ea05e3a AB |
4611 | ret = copy_to_user(arg, sa, sizeof(*sa)); |
4612 | if (ret) | |
4613 | ret = -EFAULT; | |
4614 | ||
4615 | out: | |
4616 | kfree(sa); | |
8ea05e3a AB |
4617 | return ret; |
4618 | } | |
4619 | ||
867ab667 | 4620 | static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) |
4621 | { | |
6d0379ec | 4622 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
a1b83ac5 | 4623 | size_t len; |
867ab667 | 4624 | int ret; |
a1b83ac5 AJ |
4625 | char label[BTRFS_LABEL_SIZE]; |
4626 | ||
4627 | spin_lock(&root->fs_info->super_lock); | |
4628 | memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE); | |
4629 | spin_unlock(&root->fs_info->super_lock); | |
4630 | ||
4631 | len = strnlen(label, BTRFS_LABEL_SIZE); | |
867ab667 | 4632 | |
4633 | if (len == BTRFS_LABEL_SIZE) { | |
efe120a0 FH |
4634 | btrfs_warn(root->fs_info, |
4635 | "label is too long, return the first %zu bytes", --len); | |
867ab667 | 4636 | } |
4637 | ||
867ab667 | 4638 | ret = copy_to_user(arg, label, len); |
867ab667 | 4639 | |
4640 | return ret ? -EFAULT : 0; | |
4641 | } | |
4642 | ||
a8bfd4ab | 4643 | static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) |
4644 | { | |
6d0379ec | 4645 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
a8bfd4ab | 4646 | struct btrfs_super_block *super_block = root->fs_info->super_copy; |
4647 | struct btrfs_trans_handle *trans; | |
4648 | char label[BTRFS_LABEL_SIZE]; | |
4649 | int ret; | |
4650 | ||
4651 | if (!capable(CAP_SYS_ADMIN)) | |
4652 | return -EPERM; | |
4653 | ||
4654 | if (copy_from_user(label, arg, sizeof(label))) | |
4655 | return -EFAULT; | |
4656 | ||
4657 | if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { | |
efe120a0 | 4658 | btrfs_err(root->fs_info, "unable to set label with more than %d bytes", |
a8bfd4ab | 4659 | BTRFS_LABEL_SIZE - 1); |
4660 | return -EINVAL; | |
4661 | } | |
4662 | ||
4663 | ret = mnt_want_write_file(file); | |
4664 | if (ret) | |
4665 | return ret; | |
4666 | ||
a8bfd4ab | 4667 | trans = btrfs_start_transaction(root, 0); |
4668 | if (IS_ERR(trans)) { | |
4669 | ret = PTR_ERR(trans); | |
4670 | goto out_unlock; | |
4671 | } | |
4672 | ||
a1b83ac5 | 4673 | spin_lock(&root->fs_info->super_lock); |
a8bfd4ab | 4674 | strcpy(super_block->label, label); |
a1b83ac5 | 4675 | spin_unlock(&root->fs_info->super_lock); |
d0270aca | 4676 | ret = btrfs_commit_transaction(trans, root); |
a8bfd4ab | 4677 | |
4678 | out_unlock: | |
a8bfd4ab | 4679 | mnt_drop_write_file(file); |
4680 | return ret; | |
4681 | } | |
4682 | ||
2eaa055f JM |
4683 | #define INIT_FEATURE_FLAGS(suffix) \ |
4684 | { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ | |
4685 | .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ | |
4686 | .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } | |
4687 | ||
4688 | static int btrfs_ioctl_get_supported_features(struct file *file, | |
4689 | void __user *arg) | |
4690 | { | |
4691 | static struct btrfs_ioctl_feature_flags features[3] = { | |
4692 | INIT_FEATURE_FLAGS(SUPP), | |
4693 | INIT_FEATURE_FLAGS(SAFE_SET), | |
4694 | INIT_FEATURE_FLAGS(SAFE_CLEAR) | |
4695 | }; | |
4696 | ||
4697 | if (copy_to_user(arg, &features, sizeof(features))) | |
4698 | return -EFAULT; | |
4699 | ||
4700 | return 0; | |
4701 | } | |
4702 | ||
4703 | static int btrfs_ioctl_get_features(struct file *file, void __user *arg) | |
4704 | { | |
4705 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; | |
4706 | struct btrfs_super_block *super_block = root->fs_info->super_copy; | |
4707 | struct btrfs_ioctl_feature_flags features; | |
4708 | ||
4709 | features.compat_flags = btrfs_super_compat_flags(super_block); | |
4710 | features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); | |
4711 | features.incompat_flags = btrfs_super_incompat_flags(super_block); | |
4712 | ||
4713 | if (copy_to_user(arg, &features, sizeof(features))) | |
4714 | return -EFAULT; | |
4715 | ||
4716 | return 0; | |
4717 | } | |
4718 | ||
3b02a68a JM |
4719 | static int check_feature_bits(struct btrfs_root *root, |
4720 | enum btrfs_feature_set set, | |
2eaa055f JM |
4721 | u64 change_mask, u64 flags, u64 supported_flags, |
4722 | u64 safe_set, u64 safe_clear) | |
4723 | { | |
3b02a68a JM |
4724 | const char *type = btrfs_feature_set_names[set]; |
4725 | char *names; | |
2eaa055f JM |
4726 | u64 disallowed, unsupported; |
4727 | u64 set_mask = flags & change_mask; | |
4728 | u64 clear_mask = ~flags & change_mask; | |
4729 | ||
4730 | unsupported = set_mask & ~supported_flags; | |
4731 | if (unsupported) { | |
3b02a68a JM |
4732 | names = btrfs_printable_features(set, unsupported); |
4733 | if (names) { | |
4734 | btrfs_warn(root->fs_info, | |
4735 | "this kernel does not support the %s feature bit%s", | |
4736 | names, strchr(names, ',') ? "s" : ""); | |
4737 | kfree(names); | |
4738 | } else | |
4739 | btrfs_warn(root->fs_info, | |
2eaa055f JM |
4740 | "this kernel does not support %s bits 0x%llx", |
4741 | type, unsupported); | |
4742 | return -EOPNOTSUPP; | |
4743 | } | |
4744 | ||
4745 | disallowed = set_mask & ~safe_set; | |
4746 | if (disallowed) { | |
3b02a68a JM |
4747 | names = btrfs_printable_features(set, disallowed); |
4748 | if (names) { | |
4749 | btrfs_warn(root->fs_info, | |
4750 | "can't set the %s feature bit%s while mounted", | |
4751 | names, strchr(names, ',') ? "s" : ""); | |
4752 | kfree(names); | |
4753 | } else | |
4754 | btrfs_warn(root->fs_info, | |
2eaa055f JM |
4755 | "can't set %s bits 0x%llx while mounted", |
4756 | type, disallowed); | |
4757 | return -EPERM; | |
4758 | } | |
4759 | ||
4760 | disallowed = clear_mask & ~safe_clear; | |
4761 | if (disallowed) { | |
3b02a68a JM |
4762 | names = btrfs_printable_features(set, disallowed); |
4763 | if (names) { | |
4764 | btrfs_warn(root->fs_info, | |
4765 | "can't clear the %s feature bit%s while mounted", | |
4766 | names, strchr(names, ',') ? "s" : ""); | |
4767 | kfree(names); | |
4768 | } else | |
4769 | btrfs_warn(root->fs_info, | |
2eaa055f JM |
4770 | "can't clear %s bits 0x%llx while mounted", |
4771 | type, disallowed); | |
4772 | return -EPERM; | |
4773 | } | |
4774 | ||
4775 | return 0; | |
4776 | } | |
4777 | ||
4778 | #define check_feature(root, change_mask, flags, mask_base) \ | |
3b02a68a | 4779 | check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \ |
2eaa055f JM |
4780 | BTRFS_FEATURE_ ## mask_base ## _SUPP, \ |
4781 | BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ | |
4782 | BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) | |
4783 | ||
4784 | static int btrfs_ioctl_set_features(struct file *file, void __user *arg) | |
4785 | { | |
4786 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; | |
4787 | struct btrfs_super_block *super_block = root->fs_info->super_copy; | |
4788 | struct btrfs_ioctl_feature_flags flags[2]; | |
4789 | struct btrfs_trans_handle *trans; | |
4790 | u64 newflags; | |
4791 | int ret; | |
4792 | ||
4793 | if (!capable(CAP_SYS_ADMIN)) | |
4794 | return -EPERM; | |
4795 | ||
4796 | if (copy_from_user(flags, arg, sizeof(flags))) | |
4797 | return -EFAULT; | |
4798 | ||
4799 | /* Nothing to do */ | |
4800 | if (!flags[0].compat_flags && !flags[0].compat_ro_flags && | |
4801 | !flags[0].incompat_flags) | |
4802 | return 0; | |
4803 | ||
4804 | ret = check_feature(root, flags[0].compat_flags, | |
4805 | flags[1].compat_flags, COMPAT); | |
4806 | if (ret) | |
4807 | return ret; | |
4808 | ||
4809 | ret = check_feature(root, flags[0].compat_ro_flags, | |
4810 | flags[1].compat_ro_flags, COMPAT_RO); | |
4811 | if (ret) | |
4812 | return ret; | |
4813 | ||
4814 | ret = check_feature(root, flags[0].incompat_flags, | |
4815 | flags[1].incompat_flags, INCOMPAT); | |
4816 | if (ret) | |
4817 | return ret; | |
4818 | ||
8051aa1a | 4819 | trans = btrfs_start_transaction(root, 0); |
2eaa055f JM |
4820 | if (IS_ERR(trans)) |
4821 | return PTR_ERR(trans); | |
4822 | ||
4823 | spin_lock(&root->fs_info->super_lock); | |
4824 | newflags = btrfs_super_compat_flags(super_block); | |
4825 | newflags |= flags[0].compat_flags & flags[1].compat_flags; | |
4826 | newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); | |
4827 | btrfs_set_super_compat_flags(super_block, newflags); | |
4828 | ||
4829 | newflags = btrfs_super_compat_ro_flags(super_block); | |
4830 | newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; | |
4831 | newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); | |
4832 | btrfs_set_super_compat_ro_flags(super_block, newflags); | |
4833 | ||
4834 | newflags = btrfs_super_incompat_flags(super_block); | |
4835 | newflags |= flags[0].incompat_flags & flags[1].incompat_flags; | |
4836 | newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); | |
4837 | btrfs_set_super_incompat_flags(super_block, newflags); | |
4838 | spin_unlock(&root->fs_info->super_lock); | |
4839 | ||
d0270aca | 4840 | return btrfs_commit_transaction(trans, root); |
2eaa055f JM |
4841 | } |
4842 | ||
f46b5a66 CH |
4843 | long btrfs_ioctl(struct file *file, unsigned int |
4844 | cmd, unsigned long arg) | |
4845 | { | |
496ad9aa | 4846 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
4bcabaa3 | 4847 | void __user *argp = (void __user *)arg; |
f46b5a66 CH |
4848 | |
4849 | switch (cmd) { | |
6cbff00f CH |
4850 | case FS_IOC_GETFLAGS: |
4851 | return btrfs_ioctl_getflags(file, argp); | |
4852 | case FS_IOC_SETFLAGS: | |
4853 | return btrfs_ioctl_setflags(file, argp); | |
4854 | case FS_IOC_GETVERSION: | |
4855 | return btrfs_ioctl_getversion(file, argp); | |
f7039b1d LD |
4856 | case FITRIM: |
4857 | return btrfs_ioctl_fitrim(file, argp); | |
f46b5a66 | 4858 | case BTRFS_IOC_SNAP_CREATE: |
fa0d2b9b | 4859 | return btrfs_ioctl_snap_create(file, argp, 0); |
fdfb1e4f | 4860 | case BTRFS_IOC_SNAP_CREATE_V2: |
fa0d2b9b | 4861 | return btrfs_ioctl_snap_create_v2(file, argp, 0); |
3de4586c | 4862 | case BTRFS_IOC_SUBVOL_CREATE: |
fa0d2b9b | 4863 | return btrfs_ioctl_snap_create(file, argp, 1); |
6f72c7e2 AJ |
4864 | case BTRFS_IOC_SUBVOL_CREATE_V2: |
4865 | return btrfs_ioctl_snap_create_v2(file, argp, 1); | |
76dda93c YZ |
4866 | case BTRFS_IOC_SNAP_DESTROY: |
4867 | return btrfs_ioctl_snap_destroy(file, argp); | |
0caa102d LZ |
4868 | case BTRFS_IOC_SUBVOL_GETFLAGS: |
4869 | return btrfs_ioctl_subvol_getflags(file, argp); | |
4870 | case BTRFS_IOC_SUBVOL_SETFLAGS: | |
4871 | return btrfs_ioctl_subvol_setflags(file, argp); | |
6ef5ed0d JB |
4872 | case BTRFS_IOC_DEFAULT_SUBVOL: |
4873 | return btrfs_ioctl_default_subvol(file, argp); | |
f46b5a66 | 4874 | case BTRFS_IOC_DEFRAG: |
1e701a32 CM |
4875 | return btrfs_ioctl_defrag(file, NULL); |
4876 | case BTRFS_IOC_DEFRAG_RANGE: | |
4877 | return btrfs_ioctl_defrag(file, argp); | |
f46b5a66 | 4878 | case BTRFS_IOC_RESIZE: |
198605a8 | 4879 | return btrfs_ioctl_resize(file, argp); |
f46b5a66 | 4880 | case BTRFS_IOC_ADD_DEV: |
4bcabaa3 | 4881 | return btrfs_ioctl_add_dev(root, argp); |
f46b5a66 | 4882 | case BTRFS_IOC_RM_DEV: |
da24927b | 4883 | return btrfs_ioctl_rm_dev(file, argp); |
475f6387 JS |
4884 | case BTRFS_IOC_FS_INFO: |
4885 | return btrfs_ioctl_fs_info(root, argp); | |
4886 | case BTRFS_IOC_DEV_INFO: | |
4887 | return btrfs_ioctl_dev_info(root, argp); | |
f46b5a66 | 4888 | case BTRFS_IOC_BALANCE: |
9ba1f6e4 | 4889 | return btrfs_ioctl_balance(file, NULL); |
f46b5a66 | 4890 | case BTRFS_IOC_CLONE: |
c5c9cd4d SW |
4891 | return btrfs_ioctl_clone(file, arg, 0, 0, 0); |
4892 | case BTRFS_IOC_CLONE_RANGE: | |
7a865e8a | 4893 | return btrfs_ioctl_clone_range(file, argp); |
f46b5a66 CH |
4894 | case BTRFS_IOC_TRANS_START: |
4895 | return btrfs_ioctl_trans_start(file); | |
4896 | case BTRFS_IOC_TRANS_END: | |
4897 | return btrfs_ioctl_trans_end(file); | |
ac8e9819 CM |
4898 | case BTRFS_IOC_TREE_SEARCH: |
4899 | return btrfs_ioctl_tree_search(file, argp); | |
4900 | case BTRFS_IOC_INO_LOOKUP: | |
4901 | return btrfs_ioctl_ino_lookup(file, argp); | |
d7728c96 JS |
4902 | case BTRFS_IOC_INO_PATHS: |
4903 | return btrfs_ioctl_ino_to_path(root, argp); | |
4904 | case BTRFS_IOC_LOGICAL_INO: | |
4905 | return btrfs_ioctl_logical_to_ino(root, argp); | |
1406e432 JB |
4906 | case BTRFS_IOC_SPACE_INFO: |
4907 | return btrfs_ioctl_space_info(root, argp); | |
9b199859 FDBM |
4908 | case BTRFS_IOC_SYNC: { |
4909 | int ret; | |
4910 | ||
6c255e67 | 4911 | ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1); |
9b199859 FDBM |
4912 | if (ret) |
4913 | return ret; | |
4914 | ret = btrfs_sync_fs(file->f_dentry->d_sb, 1); | |
4915 | return ret; | |
4916 | } | |
46204592 | 4917 | case BTRFS_IOC_START_SYNC: |
9a8c28be | 4918 | return btrfs_ioctl_start_sync(root, argp); |
46204592 | 4919 | case BTRFS_IOC_WAIT_SYNC: |
9a8c28be | 4920 | return btrfs_ioctl_wait_sync(root, argp); |
475f6387 | 4921 | case BTRFS_IOC_SCRUB: |
b8e95489 | 4922 | return btrfs_ioctl_scrub(file, argp); |
475f6387 JS |
4923 | case BTRFS_IOC_SCRUB_CANCEL: |
4924 | return btrfs_ioctl_scrub_cancel(root, argp); | |
4925 | case BTRFS_IOC_SCRUB_PROGRESS: | |
4926 | return btrfs_ioctl_scrub_progress(root, argp); | |
c9e9f97b | 4927 | case BTRFS_IOC_BALANCE_V2: |
9ba1f6e4 | 4928 | return btrfs_ioctl_balance(file, argp); |
837d5b6e ID |
4929 | case BTRFS_IOC_BALANCE_CTL: |
4930 | return btrfs_ioctl_balance_ctl(root, arg); | |
19a39dce ID |
4931 | case BTRFS_IOC_BALANCE_PROGRESS: |
4932 | return btrfs_ioctl_balance_progress(root, argp); | |
8ea05e3a AB |
4933 | case BTRFS_IOC_SET_RECEIVED_SUBVOL: |
4934 | return btrfs_ioctl_set_received_subvol(file, argp); | |
abccd00f HM |
4935 | #ifdef CONFIG_64BIT |
4936 | case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: | |
4937 | return btrfs_ioctl_set_received_subvol_32(file, argp); | |
4938 | #endif | |
31db9f7c AB |
4939 | case BTRFS_IOC_SEND: |
4940 | return btrfs_ioctl_send(file, argp); | |
c11d2c23 | 4941 | case BTRFS_IOC_GET_DEV_STATS: |
b27f7c0c | 4942 | return btrfs_ioctl_get_dev_stats(root, argp); |
5d13a37b | 4943 | case BTRFS_IOC_QUOTA_CTL: |
905b0dda | 4944 | return btrfs_ioctl_quota_ctl(file, argp); |
5d13a37b | 4945 | case BTRFS_IOC_QGROUP_ASSIGN: |
905b0dda | 4946 | return btrfs_ioctl_qgroup_assign(file, argp); |
5d13a37b | 4947 | case BTRFS_IOC_QGROUP_CREATE: |
905b0dda | 4948 | return btrfs_ioctl_qgroup_create(file, argp); |
5d13a37b | 4949 | case BTRFS_IOC_QGROUP_LIMIT: |
905b0dda | 4950 | return btrfs_ioctl_qgroup_limit(file, argp); |
2f232036 JS |
4951 | case BTRFS_IOC_QUOTA_RESCAN: |
4952 | return btrfs_ioctl_quota_rescan(file, argp); | |
4953 | case BTRFS_IOC_QUOTA_RESCAN_STATUS: | |
4954 | return btrfs_ioctl_quota_rescan_status(file, argp); | |
57254b6e JS |
4955 | case BTRFS_IOC_QUOTA_RESCAN_WAIT: |
4956 | return btrfs_ioctl_quota_rescan_wait(file, argp); | |
3f6bcfbd SB |
4957 | case BTRFS_IOC_DEV_REPLACE: |
4958 | return btrfs_ioctl_dev_replace(root, argp); | |
867ab667 | 4959 | case BTRFS_IOC_GET_FSLABEL: |
4960 | return btrfs_ioctl_get_fslabel(file, argp); | |
a8bfd4ab | 4961 | case BTRFS_IOC_SET_FSLABEL: |
4962 | return btrfs_ioctl_set_fslabel(file, argp); | |
416161db MF |
4963 | case BTRFS_IOC_FILE_EXTENT_SAME: |
4964 | return btrfs_ioctl_file_extent_same(file, argp); | |
2eaa055f JM |
4965 | case BTRFS_IOC_GET_SUPPORTED_FEATURES: |
4966 | return btrfs_ioctl_get_supported_features(file, argp); | |
4967 | case BTRFS_IOC_GET_FEATURES: | |
4968 | return btrfs_ioctl_get_features(file, argp); | |
4969 | case BTRFS_IOC_SET_FEATURES: | |
4970 | return btrfs_ioctl_set_features(file, argp); | |
f46b5a66 CH |
4971 | } |
4972 | ||
4973 | return -ENOTTY; | |
4974 | } |