]>
Commit | Line | Data |
---|---|---|
c1d7c514 | 1 | // SPDX-License-Identifier: GPL-2.0 |
f46b5a66 CH |
2 | /* |
3 | * Copyright (C) 2007 Oracle. All rights reserved. | |
f46b5a66 CH |
4 | */ |
5 | ||
6 | #include <linux/kernel.h> | |
7 | #include <linux/bio.h> | |
f46b5a66 CH |
8 | #include <linux/file.h> |
9 | #include <linux/fs.h> | |
cb8e7090 | 10 | #include <linux/fsnotify.h> |
f46b5a66 CH |
11 | #include <linux/pagemap.h> |
12 | #include <linux/highmem.h> | |
13 | #include <linux/time.h> | |
f46b5a66 | 14 | #include <linux/string.h> |
f46b5a66 | 15 | #include <linux/backing-dev.h> |
cb8e7090 | 16 | #include <linux/mount.h> |
cb8e7090 | 17 | #include <linux/namei.h> |
f46b5a66 | 18 | #include <linux/writeback.h> |
f46b5a66 | 19 | #include <linux/compat.h> |
cb8e7090 | 20 | #include <linux/security.h> |
f46b5a66 | 21 | #include <linux/xattr.h> |
f54de068 | 22 | #include <linux/mm.h> |
5a0e3ad6 | 23 | #include <linux/slab.h> |
f7039b1d | 24 | #include <linux/blkdev.h> |
8ea05e3a | 25 | #include <linux/uuid.h> |
55e301fd | 26 | #include <linux/btrfs.h> |
416161db | 27 | #include <linux/uaccess.h> |
ae5e165d | 28 | #include <linux/iversion.h> |
97fc2977 | 29 | #include <linux/fileattr.h> |
14605409 | 30 | #include <linux/fsverity.h> |
1881fba8 | 31 | #include <linux/sched/xacct.h> |
f46b5a66 CH |
32 | #include "ctree.h" |
33 | #include "disk-io.h" | |
949964c9 | 34 | #include "export.h" |
f46b5a66 CH |
35 | #include "transaction.h" |
36 | #include "btrfs_inode.h" | |
f46b5a66 CH |
37 | #include "print-tree.h" |
38 | #include "volumes.h" | |
925baedd | 39 | #include "locking.h" |
d7728c96 | 40 | #include "backref.h" |
606686ee | 41 | #include "rcu-string.h" |
31db9f7c | 42 | #include "send.h" |
3f6bcfbd | 43 | #include "dev-replace.h" |
63541927 | 44 | #include "props.h" |
3b02a68a | 45 | #include "sysfs.h" |
fcebe456 | 46 | #include "qgroup.h" |
1ec9a1ae | 47 | #include "tree-log.h" |
ebb8765b | 48 | #include "compression.h" |
8719aaae | 49 | #include "space-info.h" |
86736342 | 50 | #include "delalloc-space.h" |
aac0023c | 51 | #include "block-group.h" |
22b398ee | 52 | #include "subpage.h" |
f46b5a66 | 53 | |
abccd00f HM |
54 | #ifdef CONFIG_64BIT |
55 | /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI | |
56 | * structures are incorrect, as the timespec structure from userspace | |
57 | * is 4 bytes too small. We define these alternatives here to teach | |
58 | * the kernel about the 32-bit struct packing. | |
59 | */ | |
60 | struct btrfs_ioctl_timespec_32 { | |
61 | __u64 sec; | |
62 | __u32 nsec; | |
63 | } __attribute__ ((__packed__)); | |
64 | ||
65 | struct btrfs_ioctl_received_subvol_args_32 { | |
66 | char uuid[BTRFS_UUID_SIZE]; /* in */ | |
67 | __u64 stransid; /* in */ | |
68 | __u64 rtransid; /* out */ | |
69 | struct btrfs_ioctl_timespec_32 stime; /* in */ | |
70 | struct btrfs_ioctl_timespec_32 rtime; /* out */ | |
71 | __u64 flags; /* in */ | |
72 | __u64 reserved[16]; /* in */ | |
73 | } __attribute__ ((__packed__)); | |
74 | ||
75 | #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ | |
76 | struct btrfs_ioctl_received_subvol_args_32) | |
77 | #endif | |
78 | ||
2351f431 JB |
79 | #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) |
80 | struct btrfs_ioctl_send_args_32 { | |
81 | __s64 send_fd; /* in */ | |
82 | __u64 clone_sources_count; /* in */ | |
83 | compat_uptr_t clone_sources; /* in */ | |
84 | __u64 parent_root; /* in */ | |
85 | __u64 flags; /* in */ | |
e77fbf99 DS |
86 | __u32 version; /* in */ |
87 | __u8 reserved[28]; /* in */ | |
2351f431 JB |
88 | } __attribute__ ((__packed__)); |
89 | ||
90 | #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \ | |
91 | struct btrfs_ioctl_send_args_32) | |
1881fba8 OS |
92 | |
93 | struct btrfs_ioctl_encoded_io_args_32 { | |
94 | compat_uptr_t iov; | |
95 | compat_ulong_t iovcnt; | |
96 | __s64 offset; | |
97 | __u64 flags; | |
98 | __u64 len; | |
99 | __u64 unencoded_len; | |
100 | __u64 unencoded_offset; | |
101 | __u32 compression; | |
102 | __u32 encryption; | |
103 | __u8 reserved[64]; | |
104 | }; | |
105 | ||
106 | #define BTRFS_IOC_ENCODED_READ_32 _IOR(BTRFS_IOCTL_MAGIC, 64, \ | |
107 | struct btrfs_ioctl_encoded_io_args_32) | |
2351f431 | 108 | #endif |
abccd00f | 109 | |
6cbff00f | 110 | /* Mask out flags that are inappropriate for the given type of inode. */ |
1905a0f7 DS |
111 | static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode, |
112 | unsigned int flags) | |
6cbff00f | 113 | { |
1905a0f7 | 114 | if (S_ISDIR(inode->i_mode)) |
6cbff00f | 115 | return flags; |
1905a0f7 | 116 | else if (S_ISREG(inode->i_mode)) |
6cbff00f CH |
117 | return flags & ~FS_DIRSYNC_FL; |
118 | else | |
119 | return flags & (FS_NODUMP_FL | FS_NOATIME_FL); | |
120 | } | |
121 | ||
122 | /* | |
a157d4fd DS |
123 | * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS |
124 | * ioctl. | |
6cbff00f | 125 | */ |
77eea05e | 126 | static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode) |
6cbff00f CH |
127 | { |
128 | unsigned int iflags = 0; | |
77eea05e | 129 | u32 flags = binode->flags; |
14605409 | 130 | u32 ro_flags = binode->ro_flags; |
6cbff00f CH |
131 | |
132 | if (flags & BTRFS_INODE_SYNC) | |
133 | iflags |= FS_SYNC_FL; | |
134 | if (flags & BTRFS_INODE_IMMUTABLE) | |
135 | iflags |= FS_IMMUTABLE_FL; | |
136 | if (flags & BTRFS_INODE_APPEND) | |
137 | iflags |= FS_APPEND_FL; | |
138 | if (flags & BTRFS_INODE_NODUMP) | |
139 | iflags |= FS_NODUMP_FL; | |
140 | if (flags & BTRFS_INODE_NOATIME) | |
141 | iflags |= FS_NOATIME_FL; | |
142 | if (flags & BTRFS_INODE_DIRSYNC) | |
143 | iflags |= FS_DIRSYNC_FL; | |
d0092bdd LZ |
144 | if (flags & BTRFS_INODE_NODATACOW) |
145 | iflags |= FS_NOCOW_FL; | |
14605409 BB |
146 | if (ro_flags & BTRFS_INODE_RO_VERITY) |
147 | iflags |= FS_VERITY_FL; | |
d0092bdd | 148 | |
13f48dc9 | 149 | if (flags & BTRFS_INODE_NOCOMPRESS) |
d0092bdd | 150 | iflags |= FS_NOCOMP_FL; |
13f48dc9 ST |
151 | else if (flags & BTRFS_INODE_COMPRESS) |
152 | iflags |= FS_COMPR_FL; | |
6cbff00f CH |
153 | |
154 | return iflags; | |
155 | } | |
156 | ||
157 | /* | |
158 | * Update inode->i_flags based on the btrfs internal flags. | |
159 | */ | |
7b6a221e | 160 | void btrfs_sync_inode_flags_to_i_flags(struct inode *inode) |
6cbff00f | 161 | { |
5c57b8b6 | 162 | struct btrfs_inode *binode = BTRFS_I(inode); |
3cc79392 | 163 | unsigned int new_fl = 0; |
6cbff00f | 164 | |
5c57b8b6 | 165 | if (binode->flags & BTRFS_INODE_SYNC) |
3cc79392 | 166 | new_fl |= S_SYNC; |
5c57b8b6 | 167 | if (binode->flags & BTRFS_INODE_IMMUTABLE) |
3cc79392 | 168 | new_fl |= S_IMMUTABLE; |
5c57b8b6 | 169 | if (binode->flags & BTRFS_INODE_APPEND) |
3cc79392 | 170 | new_fl |= S_APPEND; |
5c57b8b6 | 171 | if (binode->flags & BTRFS_INODE_NOATIME) |
3cc79392 | 172 | new_fl |= S_NOATIME; |
5c57b8b6 | 173 | if (binode->flags & BTRFS_INODE_DIRSYNC) |
3cc79392 | 174 | new_fl |= S_DIRSYNC; |
14605409 BB |
175 | if (binode->ro_flags & BTRFS_INODE_RO_VERITY) |
176 | new_fl |= S_VERITY; | |
3cc79392 FM |
177 | |
178 | set_mask_bits(&inode->i_flags, | |
14605409 BB |
179 | S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC | |
180 | S_VERITY, new_fl); | |
6cbff00f CH |
181 | } |
182 | ||
f37c563b DS |
183 | /* |
184 | * Check if @flags are a supported and valid set of FS_*_FL flags and that | |
185 | * the old and new flags are not conflicting | |
186 | */ | |
187 | static int check_fsflags(unsigned int old_flags, unsigned int flags) | |
75e7cb7f LB |
188 | { |
189 | if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ | |
190 | FS_NOATIME_FL | FS_NODUMP_FL | \ | |
191 | FS_SYNC_FL | FS_DIRSYNC_FL | \ | |
e1e8fb6a LZ |
192 | FS_NOCOMP_FL | FS_COMPR_FL | |
193 | FS_NOCOW_FL)) | |
75e7cb7f LB |
194 | return -EOPNOTSUPP; |
195 | ||
f37c563b | 196 | /* COMPR and NOCOMP on new/old are valid */ |
75e7cb7f LB |
197 | if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) |
198 | return -EINVAL; | |
199 | ||
f37c563b DS |
200 | if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL)) |
201 | return -EINVAL; | |
202 | ||
203 | /* NOCOW and compression options are mutually exclusive */ | |
204 | if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL))) | |
205 | return -EINVAL; | |
206 | if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL))) | |
207 | return -EINVAL; | |
208 | ||
75e7cb7f LB |
209 | return 0; |
210 | } | |
211 | ||
d206e9c9 NA |
212 | static int check_fsflags_compatible(struct btrfs_fs_info *fs_info, |
213 | unsigned int flags) | |
214 | { | |
215 | if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL)) | |
216 | return -EPERM; | |
217 | ||
218 | return 0; | |
219 | } | |
220 | ||
97fc2977 MS |
221 | /* |
222 | * Set flags/xflags from the internal inode flags. The remaining items of | |
223 | * fsxattr are zeroed. | |
224 | */ | |
225 | int btrfs_fileattr_get(struct dentry *dentry, struct fileattr *fa) | |
6cbff00f | 226 | { |
97fc2977 MS |
227 | struct btrfs_inode *binode = BTRFS_I(d_inode(dentry)); |
228 | ||
77eea05e | 229 | fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(binode)); |
97fc2977 MS |
230 | return 0; |
231 | } | |
232 | ||
233 | int btrfs_fileattr_set(struct user_namespace *mnt_userns, | |
234 | struct dentry *dentry, struct fileattr *fa) | |
235 | { | |
236 | struct inode *inode = d_inode(dentry); | |
0b246afa | 237 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
5c57b8b6 DS |
238 | struct btrfs_inode *binode = BTRFS_I(inode); |
239 | struct btrfs_root *root = binode->root; | |
6cbff00f | 240 | struct btrfs_trans_handle *trans; |
5aca2842 | 241 | unsigned int fsflags, old_fsflags; |
6cbff00f | 242 | int ret; |
ff9fef55 | 243 | const char *comp = NULL; |
f37c563b | 244 | u32 binode_flags; |
6cbff00f | 245 | |
b83cc969 LZ |
246 | if (btrfs_root_readonly(root)) |
247 | return -EROFS; | |
248 | ||
97fc2977 MS |
249 | if (fileattr_has_fsx(fa)) |
250 | return -EOPNOTSUPP; | |
e7848683 | 251 | |
97fc2977 | 252 | fsflags = btrfs_mask_fsflags_for_type(inode, fa->flags); |
77eea05e | 253 | old_fsflags = btrfs_inode_flags_to_fsflags(binode); |
f37c563b DS |
254 | ret = check_fsflags(old_fsflags, fsflags); |
255 | if (ret) | |
97fc2977 | 256 | return ret; |
f37c563b | 257 | |
d206e9c9 NA |
258 | ret = check_fsflags_compatible(fs_info, fsflags); |
259 | if (ret) | |
97fc2977 | 260 | return ret; |
d206e9c9 | 261 | |
f37c563b | 262 | binode_flags = binode->flags; |
5c57b8b6 | 263 | if (fsflags & FS_SYNC_FL) |
d2b8fcfe | 264 | binode_flags |= BTRFS_INODE_SYNC; |
6cbff00f | 265 | else |
d2b8fcfe | 266 | binode_flags &= ~BTRFS_INODE_SYNC; |
5c57b8b6 | 267 | if (fsflags & FS_IMMUTABLE_FL) |
d2b8fcfe | 268 | binode_flags |= BTRFS_INODE_IMMUTABLE; |
6cbff00f | 269 | else |
d2b8fcfe | 270 | binode_flags &= ~BTRFS_INODE_IMMUTABLE; |
5c57b8b6 | 271 | if (fsflags & FS_APPEND_FL) |
d2b8fcfe | 272 | binode_flags |= BTRFS_INODE_APPEND; |
6cbff00f | 273 | else |
d2b8fcfe | 274 | binode_flags &= ~BTRFS_INODE_APPEND; |
5c57b8b6 | 275 | if (fsflags & FS_NODUMP_FL) |
d2b8fcfe | 276 | binode_flags |= BTRFS_INODE_NODUMP; |
6cbff00f | 277 | else |
d2b8fcfe | 278 | binode_flags &= ~BTRFS_INODE_NODUMP; |
5c57b8b6 | 279 | if (fsflags & FS_NOATIME_FL) |
d2b8fcfe | 280 | binode_flags |= BTRFS_INODE_NOATIME; |
6cbff00f | 281 | else |
d2b8fcfe | 282 | binode_flags &= ~BTRFS_INODE_NOATIME; |
97fc2977 MS |
283 | |
284 | /* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */ | |
285 | if (!fa->flags_valid) { | |
286 | /* 1 item for the inode */ | |
287 | trans = btrfs_start_transaction(root, 1); | |
9b8a233b RH |
288 | if (IS_ERR(trans)) |
289 | return PTR_ERR(trans); | |
97fc2977 MS |
290 | goto update_flags; |
291 | } | |
292 | ||
5c57b8b6 | 293 | if (fsflags & FS_DIRSYNC_FL) |
d2b8fcfe | 294 | binode_flags |= BTRFS_INODE_DIRSYNC; |
6cbff00f | 295 | else |
d2b8fcfe | 296 | binode_flags &= ~BTRFS_INODE_DIRSYNC; |
5c57b8b6 | 297 | if (fsflags & FS_NOCOW_FL) { |
44e5194b | 298 | if (S_ISREG(inode->i_mode)) { |
7e97b8da DS |
299 | /* |
300 | * It's safe to turn csums off here, no extents exist. | |
301 | * Otherwise we want the flag to reflect the real COW | |
302 | * status of the file and will not set it. | |
303 | */ | |
304 | if (inode->i_size == 0) | |
d2b8fcfe AJ |
305 | binode_flags |= BTRFS_INODE_NODATACOW | |
306 | BTRFS_INODE_NODATASUM; | |
7e97b8da | 307 | } else { |
d2b8fcfe | 308 | binode_flags |= BTRFS_INODE_NODATACOW; |
7e97b8da DS |
309 | } |
310 | } else { | |
311 | /* | |
01327610 | 312 | * Revert back under same assumptions as above |
7e97b8da | 313 | */ |
44e5194b | 314 | if (S_ISREG(inode->i_mode)) { |
7e97b8da | 315 | if (inode->i_size == 0) |
d2b8fcfe AJ |
316 | binode_flags &= ~(BTRFS_INODE_NODATACOW | |
317 | BTRFS_INODE_NODATASUM); | |
7e97b8da | 318 | } else { |
d2b8fcfe | 319 | binode_flags &= ~BTRFS_INODE_NODATACOW; |
7e97b8da DS |
320 | } |
321 | } | |
6cbff00f | 322 | |
75e7cb7f LB |
323 | /* |
324 | * The COMPRESS flag can only be changed by users, while the NOCOMPRESS | |
325 | * flag may be changed automatically if compression code won't make | |
326 | * things smaller. | |
327 | */ | |
5c57b8b6 | 328 | if (fsflags & FS_NOCOMP_FL) { |
d2b8fcfe AJ |
329 | binode_flags &= ~BTRFS_INODE_COMPRESS; |
330 | binode_flags |= BTRFS_INODE_NOCOMPRESS; | |
5c57b8b6 | 331 | } else if (fsflags & FS_COMPR_FL) { |
63541927 | 332 | |
97fc2977 MS |
333 | if (IS_SWAPFILE(inode)) |
334 | return -ETXTBSY; | |
eede2bf3 | 335 | |
d2b8fcfe AJ |
336 | binode_flags |= BTRFS_INODE_COMPRESS; |
337 | binode_flags &= ~BTRFS_INODE_NOCOMPRESS; | |
63541927 | 338 | |
93370509 DS |
339 | comp = btrfs_compress_type2str(fs_info->compress_type); |
340 | if (!comp || comp[0] == 0) | |
341 | comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB); | |
ebcb904d | 342 | } else { |
d2b8fcfe | 343 | binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); |
75e7cb7f | 344 | } |
6cbff00f | 345 | |
ff9fef55 AJ |
346 | /* |
347 | * 1 for inode item | |
348 | * 2 for properties | |
349 | */ | |
350 | trans = btrfs_start_transaction(root, 3); | |
97fc2977 MS |
351 | if (IS_ERR(trans)) |
352 | return PTR_ERR(trans); | |
6cbff00f | 353 | |
ff9fef55 AJ |
354 | if (comp) { |
355 | ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp, | |
356 | strlen(comp), 0); | |
357 | if (ret) { | |
358 | btrfs_abort_transaction(trans, ret); | |
359 | goto out_end_trans; | |
360 | } | |
ff9fef55 AJ |
361 | } else { |
362 | ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL, | |
363 | 0, 0); | |
364 | if (ret && ret != -ENODATA) { | |
365 | btrfs_abort_transaction(trans, ret); | |
366 | goto out_end_trans; | |
367 | } | |
368 | } | |
369 | ||
97fc2977 | 370 | update_flags: |
d2b8fcfe | 371 | binode->flags = binode_flags; |
7b6a221e | 372 | btrfs_sync_inode_flags_to_i_flags(inode); |
0c4d2d95 | 373 | inode_inc_iversion(inode); |
c2050a45 | 374 | inode->i_ctime = current_time(inode); |
9a56fcd1 | 375 | ret = btrfs_update_inode(trans, root, BTRFS_I(inode)); |
6cbff00f | 376 | |
ff9fef55 | 377 | out_end_trans: |
3a45bb20 | 378 | btrfs_end_transaction(trans); |
2d4e6f6a | 379 | return ret; |
6cbff00f CH |
380 | } |
381 | ||
0d7ed32c DS |
382 | /* |
383 | * Start exclusive operation @type, return true on success | |
384 | */ | |
c3e1f96c GR |
385 | bool btrfs_exclop_start(struct btrfs_fs_info *fs_info, |
386 | enum btrfs_exclusive_operation type) | |
387 | { | |
0d7ed32c DS |
388 | bool ret = false; |
389 | ||
390 | spin_lock(&fs_info->super_lock); | |
391 | if (fs_info->exclusive_operation == BTRFS_EXCLOP_NONE) { | |
392 | fs_info->exclusive_operation = type; | |
393 | ret = true; | |
394 | } | |
395 | spin_unlock(&fs_info->super_lock); | |
396 | ||
397 | return ret; | |
c3e1f96c GR |
398 | } |
399 | ||
578bda9e DS |
400 | /* |
401 | * Conditionally allow to enter the exclusive operation in case it's compatible | |
402 | * with the running one. This must be paired with btrfs_exclop_start_unlock and | |
403 | * btrfs_exclop_finish. | |
404 | * | |
405 | * Compatibility: | |
406 | * - the same type is already running | |
621a1ee1 | 407 | * - when trying to add a device and balance has been paused |
578bda9e DS |
408 | * - not BTRFS_EXCLOP_NONE - this is intentionally incompatible and the caller |
409 | * must check the condition first that would allow none -> @type | |
410 | */ | |
411 | bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info, | |
412 | enum btrfs_exclusive_operation type) | |
413 | { | |
414 | spin_lock(&fs_info->super_lock); | |
621a1ee1 NB |
415 | if (fs_info->exclusive_operation == type || |
416 | (fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED && | |
417 | type == BTRFS_EXCLOP_DEV_ADD)) | |
578bda9e DS |
418 | return true; |
419 | ||
420 | spin_unlock(&fs_info->super_lock); | |
421 | return false; | |
422 | } | |
423 | ||
424 | void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info) | |
425 | { | |
426 | spin_unlock(&fs_info->super_lock); | |
427 | } | |
428 | ||
c3e1f96c GR |
429 | void btrfs_exclop_finish(struct btrfs_fs_info *fs_info) |
430 | { | |
0d7ed32c | 431 | spin_lock(&fs_info->super_lock); |
c3e1f96c | 432 | WRITE_ONCE(fs_info->exclusive_operation, BTRFS_EXCLOP_NONE); |
0d7ed32c | 433 | spin_unlock(&fs_info->super_lock); |
66a2823c | 434 | sysfs_notify(&fs_info->fs_devices->fsid_kobj, NULL, "exclusive_operation"); |
c3e1f96c GR |
435 | } |
436 | ||
efc0e69c NB |
437 | void btrfs_exclop_balance(struct btrfs_fs_info *fs_info, |
438 | enum btrfs_exclusive_operation op) | |
439 | { | |
440 | switch (op) { | |
441 | case BTRFS_EXCLOP_BALANCE_PAUSED: | |
442 | spin_lock(&fs_info->super_lock); | |
443 | ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE || | |
444 | fs_info->exclusive_operation == BTRFS_EXCLOP_DEV_ADD); | |
445 | fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE_PAUSED; | |
446 | spin_unlock(&fs_info->super_lock); | |
447 | break; | |
448 | case BTRFS_EXCLOP_BALANCE: | |
449 | spin_lock(&fs_info->super_lock); | |
450 | ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED); | |
451 | fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE; | |
452 | spin_unlock(&fs_info->super_lock); | |
453 | break; | |
454 | default: | |
455 | btrfs_warn(fs_info, | |
456 | "invalid exclop balance operation %d requested", op); | |
457 | } | |
458 | } | |
459 | ||
dc408ccd | 460 | static int btrfs_ioctl_getversion(struct inode *inode, int __user *arg) |
6cbff00f | 461 | { |
6cbff00f CH |
462 | return put_user(inode->i_generation, arg); |
463 | } | |
f46b5a66 | 464 | |
b929c1d8 MPS |
465 | static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info, |
466 | void __user *arg) | |
f7039b1d | 467 | { |
f7039b1d LD |
468 | struct btrfs_device *device; |
469 | struct request_queue *q; | |
470 | struct fstrim_range range; | |
471 | u64 minlen = ULLONG_MAX; | |
472 | u64 num_devices = 0; | |
473 | int ret; | |
474 | ||
475 | if (!capable(CAP_SYS_ADMIN)) | |
476 | return -EPERM; | |
477 | ||
1cb3dc3f NA |
478 | /* |
479 | * btrfs_trim_block_group() depends on space cache, which is not | |
480 | * available in zoned filesystem. So, disallow fitrim on a zoned | |
481 | * filesystem for now. | |
482 | */ | |
483 | if (btrfs_is_zoned(fs_info)) | |
484 | return -EOPNOTSUPP; | |
485 | ||
f35f06c3 FM |
486 | /* |
487 | * If the fs is mounted with nologreplay, which requires it to be | |
488 | * mounted in RO mode as well, we can not allow discard on free space | |
489 | * inside block groups, because log trees refer to extents that are not | |
490 | * pinned in a block group's free space cache (pinning the extents is | |
491 | * precisely the first phase of replaying a log tree). | |
492 | */ | |
493 | if (btrfs_test_opt(fs_info, NOLOGREPLAY)) | |
494 | return -EROFS; | |
495 | ||
1f78160c XG |
496 | rcu_read_lock(); |
497 | list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, | |
498 | dev_list) { | |
f7039b1d LD |
499 | if (!device->bdev) |
500 | continue; | |
501 | q = bdev_get_queue(device->bdev); | |
502 | if (blk_queue_discard(q)) { | |
503 | num_devices++; | |
50d0446e | 504 | minlen = min_t(u64, q->limits.discard_granularity, |
f7039b1d LD |
505 | minlen); |
506 | } | |
507 | } | |
1f78160c | 508 | rcu_read_unlock(); |
f4c697e6 | 509 | |
f7039b1d LD |
510 | if (!num_devices) |
511 | return -EOPNOTSUPP; | |
f7039b1d LD |
512 | if (copy_from_user(&range, arg, sizeof(range))) |
513 | return -EFAULT; | |
6ba9fc8e QW |
514 | |
515 | /* | |
516 | * NOTE: Don't truncate the range using super->total_bytes. Bytenr of | |
517 | * block group is in the logical address space, which can be any | |
518 | * sectorsize aligned bytenr in the range [0, U64_MAX]. | |
519 | */ | |
520 | if (range.len < fs_info->sb->s_blocksize) | |
f4c697e6 | 521 | return -EINVAL; |
f7039b1d LD |
522 | |
523 | range.minlen = max(range.minlen, minlen); | |
2ff7e61e | 524 | ret = btrfs_trim_fs(fs_info, &range); |
f7039b1d LD |
525 | if (ret < 0) |
526 | return ret; | |
527 | ||
528 | if (copy_to_user(arg, &range, sizeof(range))) | |
529 | return -EFAULT; | |
530 | ||
531 | return 0; | |
532 | } | |
533 | ||
e1f60a65 | 534 | int __pure btrfs_is_empty_uuid(u8 *uuid) |
dd5f9615 | 535 | { |
46e0f66a CM |
536 | int i; |
537 | ||
538 | for (i = 0; i < BTRFS_UUID_SIZE; i++) { | |
539 | if (uuid[i]) | |
540 | return 0; | |
541 | } | |
542 | return 1; | |
dd5f9615 SB |
543 | } |
544 | ||
4d4340c9 CB |
545 | static noinline int create_subvol(struct user_namespace *mnt_userns, |
546 | struct inode *dir, struct dentry *dentry, | |
52f75f4f | 547 | const char *name, int namelen, |
8696c533 | 548 | struct btrfs_qgroup_inherit *inherit) |
f46b5a66 | 549 | { |
0b246afa | 550 | struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); |
f46b5a66 CH |
551 | struct btrfs_trans_handle *trans; |
552 | struct btrfs_key key; | |
49a3c4d9 | 553 | struct btrfs_root_item *root_item; |
f46b5a66 CH |
554 | struct btrfs_inode_item *inode_item; |
555 | struct extent_buffer *leaf; | |
d5c12070 | 556 | struct btrfs_root *root = BTRFS_I(dir)->root; |
76dda93c | 557 | struct btrfs_root *new_root; |
d5c12070 | 558 | struct btrfs_block_rsv block_rsv; |
95582b00 | 559 | struct timespec64 cur_time = current_time(dir); |
5662344b | 560 | struct inode *inode; |
f46b5a66 | 561 | int ret; |
2dfb1e43 | 562 | dev_t anon_dev = 0; |
f46b5a66 | 563 | u64 objectid; |
3de4586c | 564 | u64 index = 0; |
f46b5a66 | 565 | |
49a3c4d9 DS |
566 | root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); |
567 | if (!root_item) | |
568 | return -ENOMEM; | |
569 | ||
543068a2 | 570 | ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid); |
2fbe8c8a | 571 | if (ret) |
49a3c4d9 | 572 | goto fail_free; |
6a912213 | 573 | |
2dfb1e43 QW |
574 | ret = get_anon_bdev(&anon_dev); |
575 | if (ret < 0) | |
576 | goto fail_free; | |
577 | ||
e09fe2d2 QW |
578 | /* |
579 | * Don't create subvolume whose level is not zero. Or qgroup will be | |
01327610 | 580 | * screwed up since it assumes subvolume qgroup's level to be 0. |
e09fe2d2 | 581 | */ |
49a3c4d9 DS |
582 | if (btrfs_qgroup_level(objectid)) { |
583 | ret = -ENOSPC; | |
584 | goto fail_free; | |
585 | } | |
e09fe2d2 | 586 | |
d5c12070 | 587 | btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); |
9ed74f2d | 588 | /* |
d5c12070 MX |
589 | * The same as the snapshot creation, please see the comment |
590 | * of create_snapshot(). | |
9ed74f2d | 591 | */ |
c4c129db | 592 | ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false); |
d5c12070 | 593 | if (ret) |
49a3c4d9 | 594 | goto fail_free; |
d5c12070 MX |
595 | |
596 | trans = btrfs_start_transaction(root, 0); | |
597 | if (IS_ERR(trans)) { | |
598 | ret = PTR_ERR(trans); | |
e85fde51 | 599 | btrfs_subvolume_release_metadata(root, &block_rsv); |
49a3c4d9 | 600 | goto fail_free; |
d5c12070 MX |
601 | } |
602 | trans->block_rsv = &block_rsv; | |
603 | trans->bytes_reserved = block_rsv.size; | |
f46b5a66 | 604 | |
a9377422 | 605 | ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit); |
6f72c7e2 AJ |
606 | if (ret) |
607 | goto fail; | |
608 | ||
9631e4cc JB |
609 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, |
610 | BTRFS_NESTING_NORMAL); | |
8e8a1e31 JB |
611 | if (IS_ERR(leaf)) { |
612 | ret = PTR_ERR(leaf); | |
613 | goto fail; | |
614 | } | |
f46b5a66 | 615 | |
f46b5a66 CH |
616 | btrfs_mark_buffer_dirty(leaf); |
617 | ||
49a3c4d9 | 618 | inode_item = &root_item->inode; |
3cae210f QW |
619 | btrfs_set_stack_inode_generation(inode_item, 1); |
620 | btrfs_set_stack_inode_size(inode_item, 3); | |
621 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 622 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 623 | fs_info->nodesize); |
3cae210f | 624 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
f46b5a66 | 625 | |
49a3c4d9 DS |
626 | btrfs_set_root_flags(root_item, 0); |
627 | btrfs_set_root_limit(root_item, 0); | |
3cae210f | 628 | btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); |
08fe4db1 | 629 | |
49a3c4d9 DS |
630 | btrfs_set_root_bytenr(root_item, leaf->start); |
631 | btrfs_set_root_generation(root_item, trans->transid); | |
632 | btrfs_set_root_level(root_item, 0); | |
633 | btrfs_set_root_refs(root_item, 1); | |
634 | btrfs_set_root_used(root_item, leaf->len); | |
635 | btrfs_set_root_last_snapshot(root_item, 0); | |
f46b5a66 | 636 | |
49a3c4d9 DS |
637 | btrfs_set_root_generation_v2(root_item, |
638 | btrfs_root_generation(root_item)); | |
807fc790 | 639 | generate_random_guid(root_item->uuid); |
49a3c4d9 DS |
640 | btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); |
641 | btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); | |
642 | root_item->ctime = root_item->otime; | |
643 | btrfs_set_root_ctransid(root_item, trans->transid); | |
644 | btrfs_set_root_otransid(root_item, trans->transid); | |
f46b5a66 | 645 | |
925baedd | 646 | btrfs_tree_unlock(leaf); |
f46b5a66 | 647 | |
69948022 | 648 | btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID); |
f46b5a66 CH |
649 | |
650 | key.objectid = objectid; | |
5d4f98a2 | 651 | key.offset = 0; |
962a298f | 652 | key.type = BTRFS_ROOT_ITEM_KEY; |
0b246afa | 653 | ret = btrfs_insert_root(trans, fs_info->tree_root, &key, |
49a3c4d9 | 654 | root_item); |
67addf29 FM |
655 | if (ret) { |
656 | /* | |
657 | * Since we don't abort the transaction in this case, free the | |
658 | * tree block so that we don't leak space and leave the | |
659 | * filesystem in an inconsistent state (an extent item in the | |
7a163608 | 660 | * extent tree with a backreference for a root that does not |
212a58fd | 661 | * exists). |
67addf29 | 662 | */ |
212a58fd FM |
663 | btrfs_tree_lock(leaf); |
664 | btrfs_clean_tree_block(leaf); | |
665 | btrfs_tree_unlock(leaf); | |
7a163608 | 666 | btrfs_free_tree_block(trans, objectid, leaf, 0, 1); |
67addf29 | 667 | free_extent_buffer(leaf); |
f46b5a66 | 668 | goto fail; |
67addf29 FM |
669 | } |
670 | ||
671 | free_extent_buffer(leaf); | |
672 | leaf = NULL; | |
f46b5a66 | 673 | |
76dda93c | 674 | key.offset = (u64)-1; |
2dfb1e43 | 675 | new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev); |
79787eaa | 676 | if (IS_ERR(new_root)) { |
2dfb1e43 | 677 | free_anon_bdev(anon_dev); |
79787eaa | 678 | ret = PTR_ERR(new_root); |
66642832 | 679 | btrfs_abort_transaction(trans, ret); |
79787eaa JM |
680 | goto fail; |
681 | } | |
2dfb1e43 QW |
682 | /* Freeing will be done in btrfs_put_root() of new_root */ |
683 | anon_dev = 0; | |
76dda93c | 684 | |
221581e4 JB |
685 | ret = btrfs_record_root_in_trans(trans, new_root); |
686 | if (ret) { | |
687 | btrfs_put_root(new_root); | |
688 | btrfs_abort_transaction(trans, ret); | |
689 | goto fail; | |
690 | } | |
76dda93c | 691 | |
4d4340c9 | 692 | ret = btrfs_create_subvol_root(trans, new_root, root, mnt_userns); |
00246528 | 693 | btrfs_put_root(new_root); |
ce598979 MF |
694 | if (ret) { |
695 | /* We potentially lose an unused inode item here */ | |
66642832 | 696 | btrfs_abort_transaction(trans, ret); |
ce598979 MF |
697 | goto fail; |
698 | } | |
699 | ||
f46b5a66 CH |
700 | /* |
701 | * insert the directory item | |
702 | */ | |
877574e2 | 703 | ret = btrfs_set_inode_index(BTRFS_I(dir), &index); |
79787eaa | 704 | if (ret) { |
66642832 | 705 | btrfs_abort_transaction(trans, ret); |
79787eaa JM |
706 | goto fail; |
707 | } | |
3de4586c | 708 | |
684572df | 709 | ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key, |
3de4586c | 710 | BTRFS_FT_DIR, index); |
79787eaa | 711 | if (ret) { |
66642832 | 712 | btrfs_abort_transaction(trans, ret); |
f46b5a66 | 713 | goto fail; |
79787eaa | 714 | } |
0660b5af | 715 | |
6ef06d27 | 716 | btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); |
9a56fcd1 | 717 | ret = btrfs_update_inode(trans, root, BTRFS_I(dir)); |
c7e54b51 JB |
718 | if (ret) { |
719 | btrfs_abort_transaction(trans, ret); | |
720 | goto fail; | |
721 | } | |
52c26179 | 722 | |
6025c19f | 723 | ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid, |
4a0cc7ca | 724 | btrfs_ino(BTRFS_I(dir)), index, name, namelen); |
c7e54b51 JB |
725 | if (ret) { |
726 | btrfs_abort_transaction(trans, ret); | |
727 | goto fail; | |
728 | } | |
f46b5a66 | 729 | |
cdb345a8 | 730 | ret = btrfs_uuid_tree_add(trans, root_item->uuid, |
6bccf3ab | 731 | BTRFS_UUID_KEY_SUBVOL, objectid); |
dd5f9615 | 732 | if (ret) |
66642832 | 733 | btrfs_abort_transaction(trans, ret); |
dd5f9615 | 734 | |
f46b5a66 | 735 | fail: |
49a3c4d9 | 736 | kfree(root_item); |
d5c12070 MX |
737 | trans->block_rsv = NULL; |
738 | trans->bytes_reserved = 0; | |
e85fde51 | 739 | btrfs_subvolume_release_metadata(root, &block_rsv); |
de6e8200 | 740 | |
1b58ae0e FM |
741 | if (ret) |
742 | btrfs_end_transaction(trans); | |
743 | else | |
744 | ret = btrfs_commit_transaction(trans); | |
1a65e24b | 745 | |
5662344b TI |
746 | if (!ret) { |
747 | inode = btrfs_lookup_dentry(dir, dentry); | |
de6e8200 LB |
748 | if (IS_ERR(inode)) |
749 | return PTR_ERR(inode); | |
5662344b TI |
750 | d_instantiate(dentry, inode); |
751 | } | |
f46b5a66 | 752 | return ret; |
49a3c4d9 DS |
753 | |
754 | fail_free: | |
2dfb1e43 QW |
755 | if (anon_dev) |
756 | free_anon_bdev(anon_dev); | |
49a3c4d9 DS |
757 | kfree(root_item); |
758 | return ret; | |
f46b5a66 CH |
759 | } |
760 | ||
e9662f70 | 761 | static int create_snapshot(struct btrfs_root *root, struct inode *dir, |
9babda9f | 762 | struct dentry *dentry, bool readonly, |
e9662f70 | 763 | struct btrfs_qgroup_inherit *inherit) |
f46b5a66 | 764 | { |
0b246afa | 765 | struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); |
2e4bfab9 | 766 | struct inode *inode; |
f46b5a66 CH |
767 | struct btrfs_pending_snapshot *pending_snapshot; |
768 | struct btrfs_trans_handle *trans; | |
2e4bfab9 | 769 | int ret; |
f46b5a66 | 770 | |
813febdb JB |
771 | /* We do not support snapshotting right now. */ |
772 | if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { | |
773 | btrfs_warn(fs_info, | |
774 | "extent tree v2 doesn't support snapshotting yet"); | |
775 | return -EOPNOTSUPP; | |
776 | } | |
777 | ||
92a7cc42 | 778 | if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state)) |
f46b5a66 CH |
779 | return -EINVAL; |
780 | ||
eede2bf3 OS |
781 | if (atomic_read(&root->nr_swapfiles)) { |
782 | btrfs_warn(fs_info, | |
783 | "cannot snapshot subvolume with active swapfile"); | |
784 | return -ETXTBSY; | |
785 | } | |
786 | ||
23269bf5 | 787 | pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); |
a1ee7362 DS |
788 | if (!pending_snapshot) |
789 | return -ENOMEM; | |
790 | ||
2dfb1e43 QW |
791 | ret = get_anon_bdev(&pending_snapshot->anon_dev); |
792 | if (ret < 0) | |
793 | goto free_pending; | |
b0c0ea63 | 794 | pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), |
23269bf5 | 795 | GFP_KERNEL); |
8546b570 DS |
796 | pending_snapshot->path = btrfs_alloc_path(); |
797 | if (!pending_snapshot->root_item || !pending_snapshot->path) { | |
b0c0ea63 DS |
798 | ret = -ENOMEM; |
799 | goto free_pending; | |
800 | } | |
801 | ||
66d8f3dd MX |
802 | btrfs_init_block_rsv(&pending_snapshot->block_rsv, |
803 | BTRFS_BLOCK_RSV_TEMP); | |
d5c12070 MX |
804 | /* |
805 | * 1 - parent dir inode | |
806 | * 2 - dir entries | |
807 | * 1 - root item | |
808 | * 2 - root ref/backref | |
809 | * 1 - root of snapshot | |
dd5f9615 | 810 | * 1 - UUID item |
d5c12070 MX |
811 | */ |
812 | ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, | |
dd5f9615 | 813 | &pending_snapshot->block_rsv, 8, |
ee3441b4 | 814 | false); |
d5c12070 | 815 | if (ret) |
c11fbb6e | 816 | goto free_pending; |
d5c12070 | 817 | |
3de4586c | 818 | pending_snapshot->dentry = dentry; |
f46b5a66 | 819 | pending_snapshot->root = root; |
b83cc969 | 820 | pending_snapshot->readonly = readonly; |
e9662f70 | 821 | pending_snapshot->dir = dir; |
8696c533 | 822 | pending_snapshot->inherit = inherit; |
a22285a6 | 823 | |
d5c12070 | 824 | trans = btrfs_start_transaction(root, 0); |
a22285a6 YZ |
825 | if (IS_ERR(trans)) { |
826 | ret = PTR_ERR(trans); | |
827 | goto fail; | |
828 | } | |
829 | ||
28b21c55 | 830 | trans->pending_snapshot = pending_snapshot; |
9babda9f NB |
831 | |
832 | ret = btrfs_commit_transaction(trans); | |
aec8030a | 833 | if (ret) |
c37b2b62 | 834 | goto fail; |
a22285a6 YZ |
835 | |
836 | ret = pending_snapshot->error; | |
837 | if (ret) | |
838 | goto fail; | |
839 | ||
d3797308 CM |
840 | ret = btrfs_orphan_cleanup(pending_snapshot->snap); |
841 | if (ret) | |
842 | goto fail; | |
843 | ||
2b0143b5 | 844 | inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); |
2e4bfab9 YZ |
845 | if (IS_ERR(inode)) { |
846 | ret = PTR_ERR(inode); | |
847 | goto fail; | |
848 | } | |
5662344b | 849 | |
2e4bfab9 YZ |
850 | d_instantiate(dentry, inode); |
851 | ret = 0; | |
2dfb1e43 | 852 | pending_snapshot->anon_dev = 0; |
2e4bfab9 | 853 | fail: |
2dfb1e43 QW |
854 | /* Prevent double freeing of anon_dev */ |
855 | if (ret && pending_snapshot->snap) | |
856 | pending_snapshot->snap->anon_dev = 0; | |
00246528 | 857 | btrfs_put_root(pending_snapshot->snap); |
e85fde51 | 858 | btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv); |
b0c0ea63 | 859 | free_pending: |
2dfb1e43 QW |
860 | if (pending_snapshot->anon_dev) |
861 | free_anon_bdev(pending_snapshot->anon_dev); | |
b0c0ea63 | 862 | kfree(pending_snapshot->root_item); |
8546b570 | 863 | btrfs_free_path(pending_snapshot->path); |
a1ee7362 DS |
864 | kfree(pending_snapshot); |
865 | ||
f46b5a66 CH |
866 | return ret; |
867 | } | |
868 | ||
4260f7c7 SW |
869 | /* copy of may_delete in fs/namei.c() |
870 | * Check whether we can remove a link victim from directory dir, check | |
871 | * whether the type of victim is right. | |
872 | * 1. We can't do it if dir is read-only (done in permission()) | |
873 | * 2. We should have write and exec permissions on dir | |
874 | * 3. We can't remove anything from append-only dir | |
875 | * 4. We can't do anything with immutable dir (done in permission()) | |
876 | * 5. If the sticky bit on dir is set we should either | |
877 | * a. be owner of dir, or | |
878 | * b. be owner of victim, or | |
879 | * c. have CAP_FOWNER capability | |
01327610 | 880 | * 6. If the victim is append-only or immutable we can't do anything with |
4260f7c7 SW |
881 | * links pointing to it. |
882 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
883 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
884 | * 9. We can't remove a root or mountpoint. | |
885 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
886 | * nfs_async_unlink(). | |
887 | */ | |
888 | ||
c4ed533b CB |
889 | static int btrfs_may_delete(struct user_namespace *mnt_userns, |
890 | struct inode *dir, struct dentry *victim, int isdir) | |
4260f7c7 SW |
891 | { |
892 | int error; | |
893 | ||
2b0143b5 | 894 | if (d_really_is_negative(victim)) |
4260f7c7 SW |
895 | return -ENOENT; |
896 | ||
2b0143b5 | 897 | BUG_ON(d_inode(victim->d_parent) != dir); |
4fa6b5ec | 898 | audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); |
4260f7c7 | 899 | |
c4ed533b | 900 | error = inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC); |
4260f7c7 SW |
901 | if (error) |
902 | return error; | |
903 | if (IS_APPEND(dir)) | |
904 | return -EPERM; | |
c4ed533b | 905 | if (check_sticky(mnt_userns, dir, d_inode(victim)) || |
ba73d987 CB |
906 | IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) || |
907 | IS_SWAPFILE(d_inode(victim))) | |
4260f7c7 SW |
908 | return -EPERM; |
909 | if (isdir) { | |
e36cb0b8 | 910 | if (!d_is_dir(victim)) |
4260f7c7 SW |
911 | return -ENOTDIR; |
912 | if (IS_ROOT(victim)) | |
913 | return -EBUSY; | |
e36cb0b8 | 914 | } else if (d_is_dir(victim)) |
4260f7c7 SW |
915 | return -EISDIR; |
916 | if (IS_DEADDIR(dir)) | |
917 | return -ENOENT; | |
918 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
919 | return -EBUSY; | |
920 | return 0; | |
921 | } | |
922 | ||
cb8e7090 | 923 | /* copy of may_create in fs/namei.c() */ |
4d4340c9 CB |
924 | static inline int btrfs_may_create(struct user_namespace *mnt_userns, |
925 | struct inode *dir, struct dentry *child) | |
cb8e7090 | 926 | { |
2b0143b5 | 927 | if (d_really_is_positive(child)) |
cb8e7090 CH |
928 | return -EEXIST; |
929 | if (IS_DEADDIR(dir)) | |
930 | return -ENOENT; | |
4d4340c9 | 931 | if (!fsuidgid_has_mapping(dir->i_sb, mnt_userns)) |
5474bf40 | 932 | return -EOVERFLOW; |
4d4340c9 | 933 | return inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC); |
cb8e7090 CH |
934 | } |
935 | ||
936 | /* | |
937 | * Create a new subvolume below @parent. This is largely modeled after | |
938 | * sys_mkdirat and vfs_mkdir, but we only do a single component lookup | |
939 | * inside this filesystem so it's quite a bit simpler. | |
940 | */ | |
92872094 | 941 | static noinline int btrfs_mksubvol(const struct path *parent, |
4d4340c9 | 942 | struct user_namespace *mnt_userns, |
52f75f4f | 943 | const char *name, int namelen, |
72fd032e | 944 | struct btrfs_root *snap_src, |
9babda9f | 945 | bool readonly, |
8696c533 | 946 | struct btrfs_qgroup_inherit *inherit) |
cb8e7090 | 947 | { |
0b246afa JM |
948 | struct inode *dir = d_inode(parent->dentry); |
949 | struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); | |
cb8e7090 CH |
950 | struct dentry *dentry; |
951 | int error; | |
952 | ||
00235411 AV |
953 | error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); |
954 | if (error == -EINTR) | |
955 | return error; | |
cb8e7090 | 956 | |
4d4340c9 | 957 | dentry = lookup_one(mnt_userns, name, parent->dentry, namelen); |
cb8e7090 CH |
958 | error = PTR_ERR(dentry); |
959 | if (IS_ERR(dentry)) | |
960 | goto out_unlock; | |
961 | ||
4d4340c9 | 962 | error = btrfs_may_create(mnt_userns, dir, dentry); |
cb8e7090 | 963 | if (error) |
a874a63e | 964 | goto out_dput; |
cb8e7090 | 965 | |
9c52057c CM |
966 | /* |
967 | * even if this name doesn't exist, we may get hash collisions. | |
968 | * check for them now when we can safely fail | |
969 | */ | |
970 | error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, | |
971 | dir->i_ino, name, | |
972 | namelen); | |
973 | if (error) | |
974 | goto out_dput; | |
975 | ||
0b246afa | 976 | down_read(&fs_info->subvol_sem); |
76dda93c YZ |
977 | |
978 | if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) | |
979 | goto out_up_read; | |
980 | ||
9babda9f NB |
981 | if (snap_src) |
982 | error = create_snapshot(snap_src, dir, dentry, readonly, inherit); | |
983 | else | |
4d4340c9 | 984 | error = create_subvol(mnt_userns, dir, dentry, name, namelen, inherit); |
9babda9f | 985 | |
76dda93c YZ |
986 | if (!error) |
987 | fsnotify_mkdir(dir, dentry); | |
988 | out_up_read: | |
0b246afa | 989 | up_read(&fs_info->subvol_sem); |
cb8e7090 CH |
990 | out_dput: |
991 | dput(dentry); | |
992 | out_unlock: | |
64708539 | 993 | btrfs_inode_unlock(dir, 0); |
cb8e7090 CH |
994 | return error; |
995 | } | |
996 | ||
c11fbb6e | 997 | static noinline int btrfs_mksnapshot(const struct path *parent, |
4d4340c9 | 998 | struct user_namespace *mnt_userns, |
c11fbb6e RK |
999 | const char *name, int namelen, |
1000 | struct btrfs_root *root, | |
1001 | bool readonly, | |
1002 | struct btrfs_qgroup_inherit *inherit) | |
1003 | { | |
1004 | int ret; | |
1005 | bool snapshot_force_cow = false; | |
1006 | ||
1007 | /* | |
1008 | * Force new buffered writes to reserve space even when NOCOW is | |
1009 | * possible. This is to avoid later writeback (running dealloc) to | |
1010 | * fallback to COW mode and unexpectedly fail with ENOSPC. | |
1011 | */ | |
1012 | btrfs_drew_read_lock(&root->snapshot_lock); | |
1013 | ||
f9baa501 | 1014 | ret = btrfs_start_delalloc_snapshot(root, false); |
c11fbb6e RK |
1015 | if (ret) |
1016 | goto out; | |
1017 | ||
1018 | /* | |
1019 | * All previous writes have started writeback in NOCOW mode, so now | |
1020 | * we force future writes to fallback to COW mode during snapshot | |
1021 | * creation. | |
1022 | */ | |
1023 | atomic_inc(&root->snapshot_force_cow); | |
1024 | snapshot_force_cow = true; | |
1025 | ||
1026 | btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); | |
1027 | ||
4d4340c9 | 1028 | ret = btrfs_mksubvol(parent, mnt_userns, name, namelen, |
c11fbb6e RK |
1029 | root, readonly, inherit); |
1030 | out: | |
1031 | if (snapshot_force_cow) | |
1032 | atomic_dec(&root->snapshot_force_cow); | |
1033 | btrfs_drew_read_unlock(&root->snapshot_lock); | |
1034 | return ret; | |
1035 | } | |
1036 | ||
d5633b0d QW |
1037 | /* |
1038 | * Defrag specific helper to get an extent map. | |
1039 | * | |
1040 | * Differences between this and btrfs_get_extent() are: | |
1041 | * | |
1042 | * - No extent_map will be added to inode->extent_tree | |
1043 | * To reduce memory usage in the long run. | |
1044 | * | |
1045 | * - Extra optimization to skip file extents older than @newer_than | |
1046 | * By using btrfs_search_forward() we can skip entire file ranges that | |
1047 | * have extents created in past transactions, because btrfs_search_forward() | |
1048 | * will not visit leaves and nodes with a generation smaller than given | |
1049 | * minimal generation threshold (@newer_than). | |
1050 | * | |
1051 | * Return valid em if we find a file extent matching the requirement. | |
1052 | * Return NULL if we can not find a file extent matching the requirement. | |
1053 | * | |
1054 | * Return ERR_PTR() for error. | |
1055 | */ | |
1056 | static struct extent_map *defrag_get_extent(struct btrfs_inode *inode, | |
1057 | u64 start, u64 newer_than) | |
1058 | { | |
1059 | struct btrfs_root *root = inode->root; | |
1060 | struct btrfs_file_extent_item *fi; | |
1061 | struct btrfs_path path = { 0 }; | |
1062 | struct extent_map *em; | |
1063 | struct btrfs_key key; | |
1064 | u64 ino = btrfs_ino(inode); | |
1065 | int ret; | |
1066 | ||
1067 | em = alloc_extent_map(); | |
1068 | if (!em) { | |
1069 | ret = -ENOMEM; | |
1070 | goto err; | |
1071 | } | |
1072 | ||
1073 | key.objectid = ino; | |
1074 | key.type = BTRFS_EXTENT_DATA_KEY; | |
1075 | key.offset = start; | |
1076 | ||
1077 | if (newer_than) { | |
1078 | ret = btrfs_search_forward(root, &key, &path, newer_than); | |
1079 | if (ret < 0) | |
1080 | goto err; | |
1081 | /* Can't find anything newer */ | |
1082 | if (ret > 0) | |
1083 | goto not_found; | |
1084 | } else { | |
1085 | ret = btrfs_search_slot(NULL, root, &key, &path, 0, 0); | |
1086 | if (ret < 0) | |
1087 | goto err; | |
1088 | } | |
1089 | if (path.slots[0] >= btrfs_header_nritems(path.nodes[0])) { | |
1090 | /* | |
1091 | * If btrfs_search_slot() makes path to point beyond nritems, | |
1092 | * we should not have an empty leaf, as this inode must at | |
1093 | * least have its INODE_ITEM. | |
1094 | */ | |
1095 | ASSERT(btrfs_header_nritems(path.nodes[0])); | |
1096 | path.slots[0] = btrfs_header_nritems(path.nodes[0]) - 1; | |
1097 | } | |
1098 | btrfs_item_key_to_cpu(path.nodes[0], &key, path.slots[0]); | |
1099 | /* Perfect match, no need to go one slot back */ | |
1100 | if (key.objectid == ino && key.type == BTRFS_EXTENT_DATA_KEY && | |
1101 | key.offset == start) | |
1102 | goto iterate; | |
1103 | ||
1104 | /* We didn't find a perfect match, needs to go one slot back */ | |
1105 | if (path.slots[0] > 0) { | |
1106 | btrfs_item_key_to_cpu(path.nodes[0], &key, path.slots[0]); | |
1107 | if (key.objectid == ino && key.type == BTRFS_EXTENT_DATA_KEY) | |
1108 | path.slots[0]--; | |
1109 | } | |
1110 | ||
1111 | iterate: | |
1112 | /* Iterate through the path to find a file extent covering @start */ | |
1113 | while (true) { | |
1114 | u64 extent_end; | |
1115 | ||
1116 | if (path.slots[0] >= btrfs_header_nritems(path.nodes[0])) | |
1117 | goto next; | |
1118 | ||
1119 | btrfs_item_key_to_cpu(path.nodes[0], &key, path.slots[0]); | |
1120 | ||
1121 | /* | |
1122 | * We may go one slot back to INODE_REF/XATTR item, then | |
1123 | * need to go forward until we reach an EXTENT_DATA. | |
1124 | * But we should still has the correct ino as key.objectid. | |
1125 | */ | |
1126 | if (WARN_ON(key.objectid < ino) || key.type < BTRFS_EXTENT_DATA_KEY) | |
1127 | goto next; | |
1128 | ||
1129 | /* It's beyond our target range, definitely not extent found */ | |
1130 | if (key.objectid > ino || key.type > BTRFS_EXTENT_DATA_KEY) | |
1131 | goto not_found; | |
1132 | ||
1133 | /* | |
1134 | * | |<- File extent ->| | |
1135 | * \- start | |
1136 | * | |
1137 | * This means there is a hole between start and key.offset. | |
1138 | */ | |
1139 | if (key.offset > start) { | |
1140 | em->start = start; | |
1141 | em->orig_start = start; | |
1142 | em->block_start = EXTENT_MAP_HOLE; | |
1143 | em->len = key.offset - start; | |
1144 | break; | |
1145 | } | |
1146 | ||
1147 | fi = btrfs_item_ptr(path.nodes[0], path.slots[0], | |
1148 | struct btrfs_file_extent_item); | |
1149 | extent_end = btrfs_file_extent_end(&path); | |
1150 | ||
1151 | /* | |
1152 | * |<- file extent ->| | | |
1153 | * \- start | |
1154 | * | |
1155 | * We haven't reached start, search next slot. | |
1156 | */ | |
1157 | if (extent_end <= start) | |
1158 | goto next; | |
1159 | ||
1160 | /* Now this extent covers @start, convert it to em */ | |
1161 | btrfs_extent_item_to_extent_map(inode, &path, fi, false, em); | |
1162 | break; | |
1163 | next: | |
1164 | ret = btrfs_next_item(root, &path); | |
1165 | if (ret < 0) | |
1166 | goto err; | |
1167 | if (ret > 0) | |
1168 | goto not_found; | |
1169 | } | |
1170 | btrfs_release_path(&path); | |
1171 | return em; | |
1172 | ||
1173 | not_found: | |
1174 | btrfs_release_path(&path); | |
1175 | free_extent_map(em); | |
1176 | return NULL; | |
1177 | ||
1178 | err: | |
1179 | btrfs_release_path(&path); | |
1180 | free_extent_map(em); | |
1181 | return ERR_PTR(ret); | |
1182 | } | |
1183 | ||
e9eec721 | 1184 | static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start, |
d5633b0d | 1185 | u64 newer_than, bool locked) |
17ce6ef8 LB |
1186 | { |
1187 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
6c282eb4 LZ |
1188 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
1189 | struct extent_map *em; | |
76068cae | 1190 | const u32 sectorsize = BTRFS_I(inode)->root->fs_info->sectorsize; |
17ce6ef8 | 1191 | |
6c282eb4 LZ |
1192 | /* |
1193 | * hopefully we have this extent in the tree already, try without | |
1194 | * the full extent lock | |
1195 | */ | |
17ce6ef8 | 1196 | read_lock(&em_tree->lock); |
76068cae | 1197 | em = lookup_extent_mapping(em_tree, start, sectorsize); |
17ce6ef8 LB |
1198 | read_unlock(&em_tree->lock); |
1199 | ||
199257a7 QW |
1200 | /* |
1201 | * We can get a merged extent, in that case, we need to re-search | |
1202 | * tree to get the original em for defrag. | |
1203 | * | |
1204 | * If @newer_than is 0 or em::generation < newer_than, we can trust | |
1205 | * this em, as either we don't care about the generation, or the | |
1206 | * merged extent map will be rejected anyway. | |
1207 | */ | |
1208 | if (em && test_bit(EXTENT_FLAG_MERGED, &em->flags) && | |
1209 | newer_than && em->generation >= newer_than) { | |
1210 | free_extent_map(em); | |
1211 | em = NULL; | |
1212 | } | |
1213 | ||
6c282eb4 | 1214 | if (!em) { |
308d9800 | 1215 | struct extent_state *cached = NULL; |
76068cae | 1216 | u64 end = start + sectorsize - 1; |
308d9800 | 1217 | |
6c282eb4 | 1218 | /* get the big lock and read metadata off disk */ |
e9eec721 QW |
1219 | if (!locked) |
1220 | lock_extent_bits(io_tree, start, end, &cached); | |
d5633b0d | 1221 | em = defrag_get_extent(BTRFS_I(inode), start, newer_than); |
e9eec721 QW |
1222 | if (!locked) |
1223 | unlock_extent_cached(io_tree, start, end, &cached); | |
6c282eb4 LZ |
1224 | |
1225 | if (IS_ERR(em)) | |
1226 | return NULL; | |
1227 | } | |
1228 | ||
1229 | return em; | |
1230 | } | |
17ce6ef8 | 1231 | |
979b25c3 QW |
1232 | static u32 get_extent_max_capacity(const struct extent_map *em) |
1233 | { | |
1234 | if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) | |
1235 | return BTRFS_MAX_COMPRESSED; | |
1236 | return BTRFS_MAX_EXTENT_SIZE; | |
1237 | } | |
1238 | ||
e9eec721 QW |
1239 | static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em, |
1240 | bool locked) | |
6c282eb4 LZ |
1241 | { |
1242 | struct extent_map *next; | |
7093f152 | 1243 | bool ret = false; |
6c282eb4 LZ |
1244 | |
1245 | /* this is the last extent */ | |
1246 | if (em->start + em->len >= i_size_read(inode)) | |
1247 | return false; | |
1248 | ||
d5633b0d QW |
1249 | /* |
1250 | * We want to check if the next extent can be merged with the current | |
1251 | * one, which can be an extent created in a past generation, so we pass | |
1252 | * a minimum generation of 0 to defrag_lookup_extent(). | |
1253 | */ | |
1254 | next = defrag_lookup_extent(inode, em->start + em->len, 0, locked); | |
7093f152 | 1255 | /* No more em or hole */ |
e9512d72 | 1256 | if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) |
7093f152 QW |
1257 | goto out; |
1258 | if (test_bit(EXTENT_FLAG_PREALLOC, &next->flags)) | |
1259 | goto out; | |
979b25c3 QW |
1260 | /* |
1261 | * If the next extent is at its max capacity, defragging current extent | |
1262 | * makes no sense, as the total number of extents won't change. | |
1263 | */ | |
1264 | if (next->len >= get_extent_max_capacity(em)) | |
1265 | goto out; | |
7093f152 QW |
1266 | ret = true; |
1267 | out: | |
6c282eb4 | 1268 | free_extent_map(next); |
17ce6ef8 LB |
1269 | return ret; |
1270 | } | |
1271 | ||
5767b50c QW |
1272 | /* |
1273 | * Prepare one page to be defragged. | |
1274 | * | |
1275 | * This will ensure: | |
1276 | * | |
1277 | * - Returned page is locked and has been set up properly. | |
1278 | * - No ordered extent exists in the page. | |
1279 | * - The page is uptodate. | |
1280 | * | |
1281 | * NOTE: Caller should also wait for page writeback after the cluster is | |
1282 | * prepared, here we don't do writeback wait for each page. | |
1283 | */ | |
1284 | static struct page *defrag_prepare_one_page(struct btrfs_inode *inode, | |
1285 | pgoff_t index) | |
940100a4 | 1286 | { |
5767b50c QW |
1287 | struct address_space *mapping = inode->vfs_inode.i_mapping; |
1288 | gfp_t mask = btrfs_alloc_write_mask(mapping); | |
1289 | u64 page_start = (u64)index << PAGE_SHIFT; | |
1290 | u64 page_end = page_start + PAGE_SIZE - 1; | |
1291 | struct extent_state *cached_state = NULL; | |
1292 | struct page *page; | |
1293 | int ret; | |
1294 | ||
1295 | again: | |
1296 | page = find_or_create_page(mapping, index, mask); | |
1297 | if (!page) | |
1298 | return ERR_PTR(-ENOMEM); | |
940100a4 CM |
1299 | |
1300 | /* | |
24bcb454 OS |
1301 | * Since we can defragment files opened read-only, we can encounter |
1302 | * transparent huge pages here (see CONFIG_READ_ONLY_THP_FOR_FS). We | |
1303 | * can't do I/O using huge pages yet, so return an error for now. | |
1304 | * Filesystem transparent huge pages are typically only used for | |
1305 | * executables that explicitly enable them, so this isn't very | |
1306 | * restrictive. | |
940100a4 | 1307 | */ |
24bcb454 OS |
1308 | if (PageCompound(page)) { |
1309 | unlock_page(page); | |
1310 | put_page(page); | |
1311 | return ERR_PTR(-ETXTBSY); | |
1312 | } | |
940100a4 | 1313 | |
5767b50c QW |
1314 | ret = set_page_extent_mapped(page); |
1315 | if (ret < 0) { | |
1316 | unlock_page(page); | |
1317 | put_page(page); | |
1318 | return ERR_PTR(ret); | |
1319 | } | |
940100a4 | 1320 | |
5767b50c QW |
1321 | /* Wait for any existing ordered extent in the range */ |
1322 | while (1) { | |
1323 | struct btrfs_ordered_extent *ordered; | |
940100a4 | 1324 | |
5767b50c QW |
1325 | lock_extent_bits(&inode->io_tree, page_start, page_end, &cached_state); |
1326 | ordered = btrfs_lookup_ordered_range(inode, page_start, PAGE_SIZE); | |
1327 | unlock_extent_cached(&inode->io_tree, page_start, page_end, | |
1328 | &cached_state); | |
1329 | if (!ordered) | |
1330 | break; | |
17ce6ef8 | 1331 | |
5767b50c QW |
1332 | unlock_page(page); |
1333 | btrfs_start_ordered_extent(ordered, 1); | |
1334 | btrfs_put_ordered_extent(ordered); | |
1335 | lock_page(page); | |
1336 | /* | |
1337 | * We unlocked the page above, so we need check if it was | |
1338 | * released or not. | |
1339 | */ | |
1340 | if (page->mapping != mapping || !PagePrivate(page)) { | |
1341 | unlock_page(page); | |
1342 | put_page(page); | |
1343 | goto again; | |
1344 | } | |
1345 | } | |
4a3560c4 | 1346 | |
940100a4 | 1347 | /* |
5767b50c QW |
1348 | * Now the page range has no ordered extent any more. Read the page to |
1349 | * make it uptodate. | |
940100a4 | 1350 | */ |
5767b50c QW |
1351 | if (!PageUptodate(page)) { |
1352 | btrfs_readpage(NULL, page); | |
1353 | lock_page(page); | |
1354 | if (page->mapping != mapping || !PagePrivate(page)) { | |
1355 | unlock_page(page); | |
1356 | put_page(page); | |
1357 | goto again; | |
1358 | } | |
1359 | if (!PageUptodate(page)) { | |
1360 | unlock_page(page); | |
1361 | put_page(page); | |
1362 | return ERR_PTR(-EIO); | |
1363 | } | |
940100a4 | 1364 | } |
5767b50c | 1365 | return page; |
940100a4 CM |
1366 | } |
1367 | ||
eb793cf8 QW |
1368 | struct defrag_target_range { |
1369 | struct list_head list; | |
1370 | u64 start; | |
1371 | u64 len; | |
1372 | }; | |
1373 | ||
4cb5300b | 1374 | /* |
eb793cf8 | 1375 | * Collect all valid target extents. |
4cb5300b | 1376 | * |
eb793cf8 QW |
1377 | * @start: file offset to lookup |
1378 | * @len: length to lookup | |
1379 | * @extent_thresh: file extent size threshold, any extent size >= this value | |
1380 | * will be ignored | |
1381 | * @newer_than: only defrag extents newer than this value | |
1382 | * @do_compress: whether the defrag is doing compression | |
1383 | * if true, @extent_thresh will be ignored and all regular | |
1384 | * file extents meeting @newer_than will be targets. | |
e9eec721 | 1385 | * @locked: if the range has already held extent lock |
eb793cf8 | 1386 | * @target_list: list of targets file extents |
4cb5300b | 1387 | */ |
eb793cf8 QW |
1388 | static int defrag_collect_targets(struct btrfs_inode *inode, |
1389 | u64 start, u64 len, u32 extent_thresh, | |
1390 | u64 newer_than, bool do_compress, | |
966d879b QW |
1391 | bool locked, struct list_head *target_list, |
1392 | u64 *last_scanned_ret) | |
f46b5a66 | 1393 | { |
966d879b | 1394 | bool last_is_target = false; |
eb793cf8 QW |
1395 | u64 cur = start; |
1396 | int ret = 0; | |
4cb5300b | 1397 | |
eb793cf8 QW |
1398 | while (cur < start + len) { |
1399 | struct extent_map *em; | |
1400 | struct defrag_target_range *new; | |
1401 | bool next_mergeable = true; | |
1402 | u64 range_len; | |
1f12bd06 | 1403 | |
966d879b | 1404 | last_is_target = false; |
d5633b0d QW |
1405 | em = defrag_lookup_extent(&inode->vfs_inode, cur, |
1406 | newer_than, locked); | |
eb793cf8 QW |
1407 | if (!em) |
1408 | break; | |
4cb5300b | 1409 | |
eb793cf8 QW |
1410 | /* Skip hole/inline/preallocated extents */ |
1411 | if (em->block_start >= EXTENT_MAP_LAST_BYTE || | |
1412 | test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) | |
1413 | goto next; | |
4cb5300b | 1414 | |
eb793cf8 QW |
1415 | /* Skip older extent */ |
1416 | if (em->generation < newer_than) | |
1417 | goto next; | |
4cb5300b | 1418 | |
0d1ffa22 QW |
1419 | /* This em is under writeback, no need to defrag */ |
1420 | if (em->generation == (u64)-1) | |
1421 | goto next; | |
1422 | ||
0cb5950f FM |
1423 | /* |
1424 | * Our start offset might be in the middle of an existing extent | |
1425 | * map, so take that into account. | |
1426 | */ | |
1427 | range_len = em->len - (cur - em->start); | |
1428 | /* | |
1429 | * If this range of the extent map is already flagged for delalloc, | |
1430 | * skip it, because: | |
1431 | * | |
1432 | * 1) We could deadlock later, when trying to reserve space for | |
1433 | * delalloc, because in case we can't immediately reserve space | |
1434 | * the flusher can start delalloc and wait for the respective | |
1435 | * ordered extents to complete. The deadlock would happen | |
1436 | * because we do the space reservation while holding the range | |
1437 | * locked, and starting writeback, or finishing an ordered | |
1438 | * extent, requires locking the range; | |
1439 | * | |
1440 | * 2) If there's delalloc there, it means there's dirty pages for | |
1441 | * which writeback has not started yet (we clean the delalloc | |
1442 | * flag when starting writeback and after creating an ordered | |
1443 | * extent). If we mark pages in an adjacent range for defrag, | |
1444 | * then we will have a larger contiguous range for delalloc, | |
1445 | * very likely resulting in a larger extent after writeback is | |
1446 | * triggered (except in a case of free space fragmentation). | |
1447 | */ | |
1448 | if (test_range_bit(&inode->io_tree, cur, cur + range_len - 1, | |
1449 | EXTENT_DELALLOC, 0, NULL)) | |
1450 | goto next; | |
1451 | ||
eb793cf8 QW |
1452 | /* |
1453 | * For do_compress case, we want to compress all valid file | |
1454 | * extents, thus no @extent_thresh or mergeable check. | |
1455 | */ | |
1456 | if (do_compress) | |
1457 | goto add; | |
1458 | ||
1459 | /* Skip too large extent */ | |
0cb5950f | 1460 | if (range_len >= extent_thresh) |
eb793cf8 QW |
1461 | goto next; |
1462 | ||
979b25c3 QW |
1463 | /* |
1464 | * Skip extents already at its max capacity, this is mostly for | |
1465 | * compressed extents, which max cap is only 128K. | |
1466 | */ | |
1467 | if (em->len >= get_extent_max_capacity(em)) | |
1468 | goto next; | |
1469 | ||
e9eec721 QW |
1470 | next_mergeable = defrag_check_next_extent(&inode->vfs_inode, em, |
1471 | locked); | |
eb793cf8 QW |
1472 | if (!next_mergeable) { |
1473 | struct defrag_target_range *last; | |
1474 | ||
1475 | /* Empty target list, no way to merge with last entry */ | |
1476 | if (list_empty(target_list)) | |
1477 | goto next; | |
1478 | last = list_entry(target_list->prev, | |
1479 | struct defrag_target_range, list); | |
1480 | /* Not mergeable with last entry */ | |
1481 | if (last->start + last->len != cur) | |
1482 | goto next; | |
1483 | ||
1484 | /* Mergeable, fall through to add it to @target_list. */ | |
32443de3 QW |
1485 | } |
1486 | ||
eb793cf8 | 1487 | add: |
966d879b | 1488 | last_is_target = true; |
eb793cf8 QW |
1489 | range_len = min(extent_map_end(em), start + len) - cur; |
1490 | /* | |
1491 | * This one is a good target, check if it can be merged into | |
1492 | * last range of the target list. | |
1493 | */ | |
1494 | if (!list_empty(target_list)) { | |
1495 | struct defrag_target_range *last; | |
1496 | ||
1497 | last = list_entry(target_list->prev, | |
1498 | struct defrag_target_range, list); | |
1499 | ASSERT(last->start + last->len <= cur); | |
1500 | if (last->start + last->len == cur) { | |
1501 | /* Mergeable, enlarge the last entry */ | |
1502 | last->len += range_len; | |
1503 | goto next; | |
1f12bd06 | 1504 | } |
eb793cf8 | 1505 | /* Fall through to allocate a new entry */ |
600a45e1 MX |
1506 | } |
1507 | ||
eb793cf8 QW |
1508 | /* Allocate new defrag_target_range */ |
1509 | new = kmalloc(sizeof(*new), GFP_NOFS); | |
1510 | if (!new) { | |
1511 | free_extent_map(em); | |
1512 | ret = -ENOMEM; | |
1513 | break; | |
4cb5300b | 1514 | } |
eb793cf8 QW |
1515 | new->start = cur; |
1516 | new->len = range_len; | |
1517 | list_add_tail(&new->list, target_list); | |
600a45e1 | 1518 | |
eb793cf8 QW |
1519 | next: |
1520 | cur = extent_map_end(em); | |
1521 | free_extent_map(em); | |
1522 | } | |
1523 | if (ret < 0) { | |
1524 | struct defrag_target_range *entry; | |
1525 | struct defrag_target_range *tmp; | |
1526 | ||
1527 | list_for_each_entry_safe(entry, tmp, target_list, list) { | |
1528 | list_del_init(&entry->list); | |
1529 | kfree(entry); | |
600a45e1 | 1530 | } |
eb793cf8 | 1531 | } |
966d879b QW |
1532 | if (!ret && last_scanned_ret) { |
1533 | /* | |
1534 | * If the last extent is not a target, the caller can skip to | |
1535 | * the end of that extent. | |
1536 | * Otherwise, we can only go the end of the specified range. | |
1537 | */ | |
1538 | if (!last_is_target) | |
1539 | *last_scanned_ret = max(cur, *last_scanned_ret); | |
1540 | else | |
1541 | *last_scanned_ret = max(start + len, *last_scanned_ret); | |
1542 | } | |
eb793cf8 QW |
1543 | return ret; |
1544 | } | |
1545 | ||
22b398ee | 1546 | #define CLUSTER_SIZE (SZ_256K) |
a55e65b8 | 1547 | static_assert(IS_ALIGNED(CLUSTER_SIZE, PAGE_SIZE)); |
22b398ee QW |
1548 | |
1549 | /* | |
1550 | * Defrag one contiguous target range. | |
1551 | * | |
1552 | * @inode: target inode | |
1553 | * @target: target range to defrag | |
1554 | * @pages: locked pages covering the defrag range | |
1555 | * @nr_pages: number of locked pages | |
1556 | * | |
1557 | * Caller should ensure: | |
1558 | * | |
1559 | * - Pages are prepared | |
1560 | * Pages should be locked, no ordered extent in the pages range, | |
1561 | * no writeback. | |
1562 | * | |
1563 | * - Extent bits are locked | |
1564 | */ | |
1565 | static int defrag_one_locked_target(struct btrfs_inode *inode, | |
1566 | struct defrag_target_range *target, | |
1567 | struct page **pages, int nr_pages, | |
1568 | struct extent_state **cached_state) | |
1569 | { | |
1570 | struct btrfs_fs_info *fs_info = inode->root->fs_info; | |
1571 | struct extent_changeset *data_reserved = NULL; | |
1572 | const u64 start = target->start; | |
1573 | const u64 len = target->len; | |
1574 | unsigned long last_index = (start + len - 1) >> PAGE_SHIFT; | |
1575 | unsigned long start_index = start >> PAGE_SHIFT; | |
1576 | unsigned long first_index = page_index(pages[0]); | |
1577 | int ret = 0; | |
1578 | int i; | |
1579 | ||
1580 | ASSERT(last_index - first_index + 1 <= nr_pages); | |
1581 | ||
1582 | ret = btrfs_delalloc_reserve_space(inode, &data_reserved, start, len); | |
1583 | if (ret < 0) | |
1584 | return ret; | |
1585 | clear_extent_bit(&inode->io_tree, start, start + len - 1, | |
1586 | EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING | | |
1587 | EXTENT_DEFRAG, 0, 0, cached_state); | |
1588 | set_extent_defrag(&inode->io_tree, start, start + len - 1, cached_state); | |
600a45e1 | 1589 | |
22b398ee QW |
1590 | /* Update the page status */ |
1591 | for (i = start_index - first_index; i <= last_index - first_index; i++) { | |
1592 | ClearPageChecked(pages[i]); | |
1593 | btrfs_page_clamp_set_dirty(fs_info, pages[i], start, len); | |
4cb5300b | 1594 | } |
22b398ee QW |
1595 | btrfs_delalloc_release_extents(inode, len); |
1596 | extent_changeset_free(data_reserved); | |
4cb5300b | 1597 | |
22b398ee QW |
1598 | return ret; |
1599 | } | |
4cb5300b | 1600 | |
e9eec721 | 1601 | static int defrag_one_range(struct btrfs_inode *inode, u64 start, u32 len, |
966d879b QW |
1602 | u32 extent_thresh, u64 newer_than, bool do_compress, |
1603 | u64 *last_scanned_ret) | |
e9eec721 QW |
1604 | { |
1605 | struct extent_state *cached_state = NULL; | |
1606 | struct defrag_target_range *entry; | |
1607 | struct defrag_target_range *tmp; | |
1608 | LIST_HEAD(target_list); | |
1609 | struct page **pages; | |
1610 | const u32 sectorsize = inode->root->fs_info->sectorsize; | |
1611 | u64 last_index = (start + len - 1) >> PAGE_SHIFT; | |
1612 | u64 start_index = start >> PAGE_SHIFT; | |
1613 | unsigned int nr_pages = last_index - start_index + 1; | |
1614 | int ret = 0; | |
1615 | int i; | |
4cb5300b | 1616 | |
e9eec721 QW |
1617 | ASSERT(nr_pages <= CLUSTER_SIZE / PAGE_SIZE); |
1618 | ASSERT(IS_ALIGNED(start, sectorsize) && IS_ALIGNED(len, sectorsize)); | |
4cb5300b | 1619 | |
e9eec721 QW |
1620 | pages = kcalloc(nr_pages, sizeof(struct page *), GFP_NOFS); |
1621 | if (!pages) | |
1622 | return -ENOMEM; | |
7f458a38 | 1623 | |
e9eec721 QW |
1624 | /* Prepare all pages */ |
1625 | for (i = 0; i < nr_pages; i++) { | |
1626 | pages[i] = defrag_prepare_one_page(inode, start_index + i); | |
1627 | if (IS_ERR(pages[i])) { | |
1628 | ret = PTR_ERR(pages[i]); | |
1629 | pages[i] = NULL; | |
1630 | goto free_pages; | |
1631 | } | |
1632 | } | |
1633 | for (i = 0; i < nr_pages; i++) | |
1634 | wait_on_page_writeback(pages[i]); | |
1635 | ||
1636 | /* Lock the pages range */ | |
1637 | lock_extent_bits(&inode->io_tree, start_index << PAGE_SHIFT, | |
1638 | (last_index << PAGE_SHIFT) + PAGE_SIZE - 1, | |
1639 | &cached_state); | |
7f458a38 | 1640 | /* |
e9eec721 QW |
1641 | * Now we have a consistent view about the extent map, re-check |
1642 | * which range really needs to be defragged. | |
1643 | * | |
1644 | * And this time we have extent locked already, pass @locked = true | |
1645 | * so that we won't relock the extent range and cause deadlock. | |
7f458a38 | 1646 | */ |
e9eec721 QW |
1647 | ret = defrag_collect_targets(inode, start, len, extent_thresh, |
1648 | newer_than, do_compress, true, | |
966d879b | 1649 | &target_list, last_scanned_ret); |
e9eec721 QW |
1650 | if (ret < 0) |
1651 | goto unlock_extent; | |
7f458a38 | 1652 | |
e9eec721 QW |
1653 | list_for_each_entry(entry, &target_list, list) { |
1654 | ret = defrag_one_locked_target(inode, entry, pages, nr_pages, | |
1655 | &cached_state); | |
1656 | if (ret < 0) | |
1657 | break; | |
1658 | } | |
1659 | ||
1660 | list_for_each_entry_safe(entry, tmp, &target_list, list) { | |
1661 | list_del_init(&entry->list); | |
1662 | kfree(entry); | |
1663 | } | |
1664 | unlock_extent: | |
1665 | unlock_extent_cached(&inode->io_tree, start_index << PAGE_SHIFT, | |
1666 | (last_index << PAGE_SHIFT) + PAGE_SIZE - 1, | |
1667 | &cached_state); | |
1668 | free_pages: | |
1669 | for (i = 0; i < nr_pages; i++) { | |
1670 | if (pages[i]) { | |
1671 | unlock_page(pages[i]); | |
1672 | put_page(pages[i]); | |
7f458a38 | 1673 | } |
7f458a38 | 1674 | } |
e9eec721 QW |
1675 | kfree(pages); |
1676 | return ret; | |
1677 | } | |
7f458a38 | 1678 | |
b18c3ab2 QW |
1679 | static int defrag_one_cluster(struct btrfs_inode *inode, |
1680 | struct file_ra_state *ra, | |
1681 | u64 start, u32 len, u32 extent_thresh, | |
1682 | u64 newer_than, bool do_compress, | |
1683 | unsigned long *sectors_defragged, | |
966d879b QW |
1684 | unsigned long max_sectors, |
1685 | u64 *last_scanned_ret) | |
b18c3ab2 QW |
1686 | { |
1687 | const u32 sectorsize = inode->root->fs_info->sectorsize; | |
1688 | struct defrag_target_range *entry; | |
1689 | struct defrag_target_range *tmp; | |
1690 | LIST_HEAD(target_list); | |
1691 | int ret; | |
4cb5300b | 1692 | |
b18c3ab2 QW |
1693 | ret = defrag_collect_targets(inode, start, len, extent_thresh, |
1694 | newer_than, do_compress, false, | |
966d879b | 1695 | &target_list, NULL); |
b18c3ab2 QW |
1696 | if (ret < 0) |
1697 | goto out; | |
4cb5300b | 1698 | |
b18c3ab2 QW |
1699 | list_for_each_entry(entry, &target_list, list) { |
1700 | u32 range_len = entry->len; | |
4cb5300b | 1701 | |
484167da | 1702 | /* Reached or beyond the limit */ |
c080b414 QW |
1703 | if (max_sectors && *sectors_defragged >= max_sectors) { |
1704 | ret = 1; | |
b18c3ab2 | 1705 | break; |
c080b414 | 1706 | } |
4cb5300b | 1707 | |
b18c3ab2 QW |
1708 | if (max_sectors) |
1709 | range_len = min_t(u32, range_len, | |
1710 | (max_sectors - *sectors_defragged) * sectorsize); | |
4cb5300b | 1711 | |
966d879b QW |
1712 | /* |
1713 | * If defrag_one_range() has updated last_scanned_ret, | |
1714 | * our range may already be invalid (e.g. hole punched). | |
1715 | * Skip if our range is before last_scanned_ret, as there is | |
1716 | * no need to defrag the range anymore. | |
1717 | */ | |
1718 | if (entry->start + range_len <= *last_scanned_ret) | |
1719 | continue; | |
1720 | ||
b18c3ab2 QW |
1721 | if (ra) |
1722 | page_cache_sync_readahead(inode->vfs_inode.i_mapping, | |
1723 | ra, NULL, entry->start >> PAGE_SHIFT, | |
1724 | ((entry->start + range_len - 1) >> PAGE_SHIFT) - | |
1725 | (entry->start >> PAGE_SHIFT) + 1); | |
1726 | /* | |
1727 | * Here we may not defrag any range if holes are punched before | |
1728 | * we locked the pages. | |
1729 | * But that's fine, it only affects the @sectors_defragged | |
1730 | * accounting. | |
1731 | */ | |
1732 | ret = defrag_one_range(inode, entry->start, range_len, | |
966d879b QW |
1733 | extent_thresh, newer_than, do_compress, |
1734 | last_scanned_ret); | |
b18c3ab2 QW |
1735 | if (ret < 0) |
1736 | break; | |
484167da QW |
1737 | *sectors_defragged += range_len >> |
1738 | inode->root->fs_info->sectorsize_bits; | |
4cb5300b | 1739 | } |
4cb5300b | 1740 | out: |
b18c3ab2 QW |
1741 | list_for_each_entry_safe(entry, tmp, &target_list, list) { |
1742 | list_del_init(&entry->list); | |
1743 | kfree(entry); | |
4cb5300b | 1744 | } |
966d879b QW |
1745 | if (ret >= 0) |
1746 | *last_scanned_ret = max(*last_scanned_ret, start + len); | |
4cb5300b | 1747 | return ret; |
4cb5300b CM |
1748 | } |
1749 | ||
1ccc2e8a QW |
1750 | /* |
1751 | * Entry point to file defragmentation. | |
1752 | * | |
1753 | * @inode: inode to be defragged | |
1754 | * @ra: readahead state (can be NUL) | |
1755 | * @range: defrag options including range and flags | |
1756 | * @newer_than: minimum transid to defrag | |
1757 | * @max_to_defrag: max number of sectors to be defragged, if 0, the whole inode | |
1758 | * will be defragged. | |
484167da QW |
1759 | * |
1760 | * Return <0 for error. | |
c080b414 QW |
1761 | * Return >=0 for the number of sectors defragged, and range->start will be updated |
1762 | * to indicate the file offset where next defrag should be started at. | |
1763 | * (Mostly for autodefrag, which sets @max_to_defrag thus we may exit early without | |
1764 | * defragging all the range). | |
1ccc2e8a QW |
1765 | */ |
1766 | int btrfs_defrag_file(struct inode *inode, struct file_ra_state *ra, | |
4cb5300b CM |
1767 | struct btrfs_ioctl_defrag_range_args *range, |
1768 | u64 newer_than, unsigned long max_to_defrag) | |
1769 | { | |
0b246afa | 1770 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
7b508037 | 1771 | unsigned long sectors_defragged = 0; |
151a31b2 | 1772 | u64 isize = i_size_read(inode); |
7b508037 QW |
1773 | u64 cur; |
1774 | u64 last_byte; | |
1e2ef46d | 1775 | bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; |
1ccc2e8a | 1776 | bool ra_allocated = false; |
1a419d85 | 1777 | int compress_type = BTRFS_COMPRESS_ZLIB; |
7b508037 | 1778 | int ret = 0; |
aab110ab | 1779 | u32 extent_thresh = range->extent_thresh; |
27cdfde1 | 1780 | pgoff_t start_index; |
4cb5300b | 1781 | |
0abd5b17 LB |
1782 | if (isize == 0) |
1783 | return 0; | |
1784 | ||
1785 | if (range->start >= isize) | |
1786 | return -EINVAL; | |
1a419d85 | 1787 | |
1e2ef46d | 1788 | if (do_compress) { |
ce96b7ff | 1789 | if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES) |
1a419d85 LZ |
1790 | return -EINVAL; |
1791 | if (range->compress_type) | |
1792 | compress_type = range->compress_type; | |
1793 | } | |
f46b5a66 | 1794 | |
0abd5b17 | 1795 | if (extent_thresh == 0) |
ee22184b | 1796 | extent_thresh = SZ_256K; |
940100a4 | 1797 | |
7b508037 QW |
1798 | if (range->start + range->len > range->start) { |
1799 | /* Got a specific range */ | |
6b34cd8e | 1800 | last_byte = min(isize, range->start + range->len); |
7b508037 QW |
1801 | } else { |
1802 | /* Defrag until file end */ | |
6b34cd8e | 1803 | last_byte = isize; |
7b508037 QW |
1804 | } |
1805 | ||
6b34cd8e FM |
1806 | /* Align the range */ |
1807 | cur = round_down(range->start, fs_info->sectorsize); | |
1808 | last_byte = round_up(last_byte, fs_info->sectorsize) - 1; | |
1809 | ||
4cb5300b | 1810 | /* |
1ccc2e8a | 1811 | * If we were not given a ra, allocate a readahead context. As |
0a52d108 DS |
1812 | * readahead is just an optimization, defrag will work without it so |
1813 | * we don't error out. | |
4cb5300b | 1814 | */ |
1ccc2e8a QW |
1815 | if (!ra) { |
1816 | ra_allocated = true; | |
63e727ec | 1817 | ra = kzalloc(sizeof(*ra), GFP_KERNEL); |
0a52d108 DS |
1818 | if (ra) |
1819 | file_ra_state_init(ra, inode->i_mapping); | |
4cb5300b | 1820 | } |
4cb5300b | 1821 | |
27cdfde1 FM |
1822 | /* |
1823 | * Make writeback start from the beginning of the range, so that the | |
1824 | * defrag range can be written sequentially. | |
1825 | */ | |
1826 | start_index = cur >> PAGE_SHIFT; | |
1827 | if (start_index < inode->i_mapping->writeback_index) | |
1828 | inode->i_mapping->writeback_index = start_index; | |
1829 | ||
7b508037 | 1830 | while (cur < last_byte) { |
3c9d31c7 | 1831 | const unsigned long prev_sectors_defragged = sectors_defragged; |
966d879b | 1832 | u64 last_scanned = cur; |
7b508037 | 1833 | u64 cluster_end; |
008873ea | 1834 | |
b767c2fc FM |
1835 | if (btrfs_defrag_cancelled(fs_info)) { |
1836 | ret = -EAGAIN; | |
1837 | break; | |
1838 | } | |
1839 | ||
7b508037 QW |
1840 | /* We want the cluster end at page boundary when possible */ |
1841 | cluster_end = (((cur >> PAGE_SHIFT) + | |
1842 | (SZ_256K >> PAGE_SHIFT)) << PAGE_SHIFT) - 1; | |
1843 | cluster_end = min(cluster_end, last_byte); | |
940100a4 | 1844 | |
64708539 | 1845 | btrfs_inode_lock(inode, 0); |
eede2bf3 OS |
1846 | if (IS_SWAPFILE(inode)) { |
1847 | ret = -ETXTBSY; | |
7b508037 QW |
1848 | btrfs_inode_unlock(inode, 0); |
1849 | break; | |
eede2bf3 | 1850 | } |
7b508037 | 1851 | if (!(inode->i_sb->s_flags & SB_ACTIVE)) { |
64708539 | 1852 | btrfs_inode_unlock(inode, 0); |
7b508037 | 1853 | break; |
ecb8bea8 | 1854 | } |
7b508037 QW |
1855 | if (do_compress) |
1856 | BTRFS_I(inode)->defrag_compress = compress_type; | |
1857 | ret = defrag_one_cluster(BTRFS_I(inode), ra, cur, | |
1858 | cluster_end + 1 - cur, extent_thresh, | |
966d879b QW |
1859 | newer_than, do_compress, §ors_defragged, |
1860 | max_to_defrag, &last_scanned); | |
3c9d31c7 FM |
1861 | |
1862 | if (sectors_defragged > prev_sectors_defragged) | |
1863 | balance_dirty_pages_ratelimited(inode->i_mapping); | |
1864 | ||
64708539 | 1865 | btrfs_inode_unlock(inode, 0); |
7b508037 QW |
1866 | if (ret < 0) |
1867 | break; | |
966d879b | 1868 | cur = max(cluster_end + 1, last_scanned); |
c080b414 QW |
1869 | if (ret > 0) { |
1870 | ret = 0; | |
1871 | break; | |
1872 | } | |
ea0eba69 | 1873 | cond_resched(); |
f46b5a66 CH |
1874 | } |
1875 | ||
7b508037 QW |
1876 | if (ra_allocated) |
1877 | kfree(ra); | |
c080b414 QW |
1878 | /* |
1879 | * Update range.start for autodefrag, this will indicate where to start | |
1880 | * in next run. | |
1881 | */ | |
1882 | range->start = cur; | |
7b508037 QW |
1883 | if (sectors_defragged) { |
1884 | /* | |
1885 | * We have defragged some sectors, for compression case they | |
1886 | * need to be written back immediately. | |
1887 | */ | |
1888 | if (range->flags & BTRFS_DEFRAG_RANGE_START_IO) { | |
dec8ef90 | 1889 | filemap_flush(inode->i_mapping); |
7b508037 QW |
1890 | if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, |
1891 | &BTRFS_I(inode)->runtime_flags)) | |
1892 | filemap_flush(inode->i_mapping); | |
1893 | } | |
1894 | if (range->compress_type == BTRFS_COMPRESS_LZO) | |
1895 | btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); | |
1896 | else if (range->compress_type == BTRFS_COMPRESS_ZSTD) | |
1897 | btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); | |
1898 | ret = sectors_defragged; | |
dec8ef90 | 1899 | } |
1e2ef46d | 1900 | if (do_compress) { |
64708539 | 1901 | btrfs_inode_lock(inode, 0); |
eec63c65 | 1902 | BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; |
64708539 | 1903 | btrfs_inode_unlock(inode, 0); |
633085c7 | 1904 | } |
940100a4 | 1905 | return ret; |
f46b5a66 CH |
1906 | } |
1907 | ||
17aaa434 DS |
1908 | /* |
1909 | * Try to start exclusive operation @type or cancel it if it's running. | |
1910 | * | |
1911 | * Return: | |
1912 | * 0 - normal mode, newly claimed op started | |
1913 | * >0 - normal mode, something else is running, | |
1914 | * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space | |
1915 | * ECANCELED - cancel mode, successful cancel | |
1916 | * ENOTCONN - cancel mode, operation not running anymore | |
1917 | */ | |
1918 | static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info, | |
1919 | enum btrfs_exclusive_operation type, bool cancel) | |
1920 | { | |
1921 | if (!cancel) { | |
1922 | /* Start normal op */ | |
1923 | if (!btrfs_exclop_start(fs_info, type)) | |
1924 | return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; | |
1925 | /* Exclusive operation is now claimed */ | |
1926 | return 0; | |
1927 | } | |
1928 | ||
1929 | /* Cancel running op */ | |
1930 | if (btrfs_exclop_start_try_lock(fs_info, type)) { | |
1931 | /* | |
1932 | * This blocks any exclop finish from setting it to NONE, so we | |
1933 | * request cancellation. Either it runs and we will wait for it, | |
1934 | * or it has finished and no waiting will happen. | |
1935 | */ | |
1936 | atomic_inc(&fs_info->reloc_cancel_req); | |
1937 | btrfs_exclop_start_unlock(fs_info); | |
1938 | ||
1939 | if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags)) | |
1940 | wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING, | |
1941 | TASK_INTERRUPTIBLE); | |
1942 | ||
1943 | return -ECANCELED; | |
1944 | } | |
1945 | ||
1946 | /* Something else is running or none */ | |
1947 | return -ENOTCONN; | |
1948 | } | |
1949 | ||
198605a8 | 1950 | static noinline int btrfs_ioctl_resize(struct file *file, |
76dda93c | 1951 | void __user *arg) |
f46b5a66 | 1952 | { |
562d7b15 | 1953 | BTRFS_DEV_LOOKUP_ARGS(args); |
0b246afa JM |
1954 | struct inode *inode = file_inode(file); |
1955 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
f46b5a66 CH |
1956 | u64 new_size; |
1957 | u64 old_size; | |
1958 | u64 devid = 1; | |
0b246afa | 1959 | struct btrfs_root *root = BTRFS_I(inode)->root; |
f46b5a66 CH |
1960 | struct btrfs_ioctl_vol_args *vol_args; |
1961 | struct btrfs_trans_handle *trans; | |
1962 | struct btrfs_device *device = NULL; | |
1963 | char *sizestr; | |
9a40f122 | 1964 | char *retptr; |
f46b5a66 CH |
1965 | char *devstr = NULL; |
1966 | int ret = 0; | |
f46b5a66 | 1967 | int mod = 0; |
bb059a37 | 1968 | bool cancel; |
f46b5a66 | 1969 | |
e441d54d CM |
1970 | if (!capable(CAP_SYS_ADMIN)) |
1971 | return -EPERM; | |
1972 | ||
198605a8 MX |
1973 | ret = mnt_want_write_file(file); |
1974 | if (ret) | |
1975 | return ret; | |
1976 | ||
bb059a37 DS |
1977 | /* |
1978 | * Read the arguments before checking exclusivity to be able to | |
1979 | * distinguish regular resize and cancel | |
1980 | */ | |
dae7b665 | 1981 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
1982 | if (IS_ERR(vol_args)) { |
1983 | ret = PTR_ERR(vol_args); | |
bb059a37 | 1984 | goto out_drop; |
c9e9f97b | 1985 | } |
5516e595 | 1986 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
f46b5a66 | 1987 | sizestr = vol_args->name; |
bb059a37 DS |
1988 | cancel = (strcmp("cancel", sizestr) == 0); |
1989 | ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel); | |
1990 | if (ret) | |
1991 | goto out_free; | |
1992 | /* Exclusive operation is now claimed */ | |
1993 | ||
f46b5a66 CH |
1994 | devstr = strchr(sizestr, ':'); |
1995 | if (devstr) { | |
f46b5a66 CH |
1996 | sizestr = devstr + 1; |
1997 | *devstr = '\0'; | |
1998 | devstr = vol_args->name; | |
58dfae63 Z |
1999 | ret = kstrtoull(devstr, 10, &devid); |
2000 | if (ret) | |
bb059a37 | 2001 | goto out_finish; |
dfd79829 MX |
2002 | if (!devid) { |
2003 | ret = -EINVAL; | |
bb059a37 | 2004 | goto out_finish; |
dfd79829 | 2005 | } |
0b246afa | 2006 | btrfs_info(fs_info, "resizing devid %llu", devid); |
f46b5a66 | 2007 | } |
dba60f3f | 2008 | |
562d7b15 JB |
2009 | args.devid = devid; |
2010 | device = btrfs_find_device(fs_info->fs_devices, &args); | |
f46b5a66 | 2011 | if (!device) { |
0b246afa JM |
2012 | btrfs_info(fs_info, "resizer unable to find device %llu", |
2013 | devid); | |
dfd79829 | 2014 | ret = -ENODEV; |
bb059a37 | 2015 | goto out_finish; |
f46b5a66 | 2016 | } |
dba60f3f | 2017 | |
ebbede42 | 2018 | if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { |
0b246afa | 2019 | btrfs_info(fs_info, |
efe120a0 | 2020 | "resizer unable to apply on readonly device %llu", |
c1c9ff7c | 2021 | devid); |
dfd79829 | 2022 | ret = -EPERM; |
bb059a37 | 2023 | goto out_finish; |
4e42ae1b LB |
2024 | } |
2025 | ||
f46b5a66 | 2026 | if (!strcmp(sizestr, "max")) |
cda00eba | 2027 | new_size = bdev_nr_bytes(device->bdev); |
f46b5a66 CH |
2028 | else { |
2029 | if (sizestr[0] == '-') { | |
2030 | mod = -1; | |
2031 | sizestr++; | |
2032 | } else if (sizestr[0] == '+') { | |
2033 | mod = 1; | |
2034 | sizestr++; | |
2035 | } | |
9a40f122 GH |
2036 | new_size = memparse(sizestr, &retptr); |
2037 | if (*retptr != '\0' || new_size == 0) { | |
f46b5a66 | 2038 | ret = -EINVAL; |
bb059a37 | 2039 | goto out_finish; |
f46b5a66 CH |
2040 | } |
2041 | } | |
2042 | ||
401e29c1 | 2043 | if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { |
dfd79829 | 2044 | ret = -EPERM; |
bb059a37 | 2045 | goto out_finish; |
63a212ab SB |
2046 | } |
2047 | ||
7cc8e58d | 2048 | old_size = btrfs_device_get_total_bytes(device); |
f46b5a66 CH |
2049 | |
2050 | if (mod < 0) { | |
2051 | if (new_size > old_size) { | |
2052 | ret = -EINVAL; | |
bb059a37 | 2053 | goto out_finish; |
f46b5a66 CH |
2054 | } |
2055 | new_size = old_size - new_size; | |
2056 | } else if (mod > 0) { | |
eb8052e0 | 2057 | if (new_size > ULLONG_MAX - old_size) { |
902c68a4 | 2058 | ret = -ERANGE; |
bb059a37 | 2059 | goto out_finish; |
eb8052e0 | 2060 | } |
f46b5a66 CH |
2061 | new_size = old_size + new_size; |
2062 | } | |
2063 | ||
ee22184b | 2064 | if (new_size < SZ_256M) { |
f46b5a66 | 2065 | ret = -EINVAL; |
bb059a37 | 2066 | goto out_finish; |
f46b5a66 | 2067 | } |
cda00eba | 2068 | if (new_size > bdev_nr_bytes(device->bdev)) { |
f46b5a66 | 2069 | ret = -EFBIG; |
bb059a37 | 2070 | goto out_finish; |
f46b5a66 CH |
2071 | } |
2072 | ||
47f08b96 | 2073 | new_size = round_down(new_size, fs_info->sectorsize); |
f46b5a66 | 2074 | |
f46b5a66 | 2075 | if (new_size > old_size) { |
a22285a6 | 2076 | trans = btrfs_start_transaction(root, 0); |
98d5dc13 TI |
2077 | if (IS_ERR(trans)) { |
2078 | ret = PTR_ERR(trans); | |
bb059a37 | 2079 | goto out_finish; |
98d5dc13 | 2080 | } |
f46b5a66 | 2081 | ret = btrfs_grow_device(trans, device, new_size); |
3a45bb20 | 2082 | btrfs_commit_transaction(trans); |
ece7d20e | 2083 | } else if (new_size < old_size) { |
f46b5a66 | 2084 | ret = btrfs_shrink_device(device, new_size); |
0253f40e | 2085 | } /* equal, nothing need to do */ |
f46b5a66 | 2086 | |
faf8f7b9 MPS |
2087 | if (ret == 0 && new_size != old_size) |
2088 | btrfs_info_in_rcu(fs_info, | |
2089 | "resize device %s (devid %llu) from %llu to %llu", | |
2090 | rcu_str_deref(device->name), device->devid, | |
2091 | old_size, new_size); | |
bb059a37 DS |
2092 | out_finish: |
2093 | btrfs_exclop_finish(fs_info); | |
c9e9f97b | 2094 | out_free: |
f46b5a66 | 2095 | kfree(vol_args); |
bb059a37 | 2096 | out_drop: |
18f39c41 | 2097 | mnt_drop_write_file(file); |
f46b5a66 CH |
2098 | return ret; |
2099 | } | |
2100 | ||
5d54c67e | 2101 | static noinline int __btrfs_ioctl_snap_create(struct file *file, |
4d4340c9 | 2102 | struct user_namespace *mnt_userns, |
52f75f4f | 2103 | const char *name, unsigned long fd, int subvol, |
5d54c67e | 2104 | bool readonly, |
8696c533 | 2105 | struct btrfs_qgroup_inherit *inherit) |
f46b5a66 | 2106 | { |
f46b5a66 | 2107 | int namelen; |
3de4586c | 2108 | int ret = 0; |
f46b5a66 | 2109 | |
325c50e3 JM |
2110 | if (!S_ISDIR(file_inode(file)->i_mode)) |
2111 | return -ENOTDIR; | |
2112 | ||
a874a63e LB |
2113 | ret = mnt_want_write_file(file); |
2114 | if (ret) | |
2115 | goto out; | |
2116 | ||
72fd032e SW |
2117 | namelen = strlen(name); |
2118 | if (strchr(name, '/')) { | |
f46b5a66 | 2119 | ret = -EINVAL; |
a874a63e | 2120 | goto out_drop_write; |
f46b5a66 CH |
2121 | } |
2122 | ||
16780cab CM |
2123 | if (name[0] == '.' && |
2124 | (namelen == 1 || (name[1] == '.' && namelen == 2))) { | |
2125 | ret = -EEXIST; | |
a874a63e | 2126 | goto out_drop_write; |
16780cab CM |
2127 | } |
2128 | ||
3de4586c | 2129 | if (subvol) { |
4d4340c9 CB |
2130 | ret = btrfs_mksubvol(&file->f_path, mnt_userns, name, |
2131 | namelen, NULL, readonly, inherit); | |
cb8e7090 | 2132 | } else { |
2903ff01 | 2133 | struct fd src = fdget(fd); |
3de4586c | 2134 | struct inode *src_inode; |
2903ff01 | 2135 | if (!src.file) { |
3de4586c | 2136 | ret = -EINVAL; |
a874a63e | 2137 | goto out_drop_write; |
3de4586c CM |
2138 | } |
2139 | ||
496ad9aa AV |
2140 | src_inode = file_inode(src.file); |
2141 | if (src_inode->i_sb != file_inode(file)->i_sb) { | |
c79b4713 | 2142 | btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, |
efe120a0 | 2143 | "Snapshot src from another FS"); |
23ad5b17 | 2144 | ret = -EXDEV; |
4d4340c9 | 2145 | } else if (!inode_owner_or_capable(mnt_userns, src_inode)) { |
d0242061 DS |
2146 | /* |
2147 | * Subvolume creation is not restricted, but snapshots | |
2148 | * are limited to own subvolumes only | |
2149 | */ | |
2150 | ret = -EPERM; | |
ecd18815 | 2151 | } else { |
4d4340c9 CB |
2152 | ret = btrfs_mksnapshot(&file->f_path, mnt_userns, |
2153 | name, namelen, | |
2154 | BTRFS_I(src_inode)->root, | |
2155 | readonly, inherit); | |
3de4586c | 2156 | } |
2903ff01 | 2157 | fdput(src); |
cb8e7090 | 2158 | } |
a874a63e LB |
2159 | out_drop_write: |
2160 | mnt_drop_write_file(file); | |
f46b5a66 | 2161 | out: |
72fd032e SW |
2162 | return ret; |
2163 | } | |
2164 | ||
2165 | static noinline int btrfs_ioctl_snap_create(struct file *file, | |
fa0d2b9b | 2166 | void __user *arg, int subvol) |
72fd032e | 2167 | { |
fa0d2b9b | 2168 | struct btrfs_ioctl_vol_args *vol_args; |
72fd032e SW |
2169 | int ret; |
2170 | ||
325c50e3 JM |
2171 | if (!S_ISDIR(file_inode(file)->i_mode)) |
2172 | return -ENOTDIR; | |
2173 | ||
fa0d2b9b LZ |
2174 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
2175 | if (IS_ERR(vol_args)) | |
2176 | return PTR_ERR(vol_args); | |
2177 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
72fd032e | 2178 | |
4d4340c9 CB |
2179 | ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file), |
2180 | vol_args->name, vol_args->fd, subvol, | |
2181 | false, NULL); | |
fdfb1e4f | 2182 | |
fa0d2b9b LZ |
2183 | kfree(vol_args); |
2184 | return ret; | |
2185 | } | |
fdfb1e4f | 2186 | |
fa0d2b9b LZ |
2187 | static noinline int btrfs_ioctl_snap_create_v2(struct file *file, |
2188 | void __user *arg, int subvol) | |
2189 | { | |
2190 | struct btrfs_ioctl_vol_args_v2 *vol_args; | |
2191 | int ret; | |
b83cc969 | 2192 | bool readonly = false; |
6f72c7e2 | 2193 | struct btrfs_qgroup_inherit *inherit = NULL; |
75eaa0e2 | 2194 | |
325c50e3 JM |
2195 | if (!S_ISDIR(file_inode(file)->i_mode)) |
2196 | return -ENOTDIR; | |
2197 | ||
fa0d2b9b LZ |
2198 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
2199 | if (IS_ERR(vol_args)) | |
2200 | return PTR_ERR(vol_args); | |
2201 | vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; | |
75eaa0e2 | 2202 | |
673990db | 2203 | if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) { |
b83cc969 | 2204 | ret = -EOPNOTSUPP; |
c47ca32d | 2205 | goto free_args; |
72fd032e | 2206 | } |
fa0d2b9b | 2207 | |
b83cc969 LZ |
2208 | if (vol_args->flags & BTRFS_SUBVOL_RDONLY) |
2209 | readonly = true; | |
6f72c7e2 | 2210 | if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { |
5011c5a6 DC |
2211 | u64 nums; |
2212 | ||
2213 | if (vol_args->size < sizeof(*inherit) || | |
2214 | vol_args->size > PAGE_SIZE) { | |
6f72c7e2 | 2215 | ret = -EINVAL; |
c47ca32d | 2216 | goto free_args; |
6f72c7e2 AJ |
2217 | } |
2218 | inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); | |
2219 | if (IS_ERR(inherit)) { | |
2220 | ret = PTR_ERR(inherit); | |
c47ca32d | 2221 | goto free_args; |
6f72c7e2 | 2222 | } |
5011c5a6 DC |
2223 | |
2224 | if (inherit->num_qgroups > PAGE_SIZE || | |
2225 | inherit->num_ref_copies > PAGE_SIZE || | |
2226 | inherit->num_excl_copies > PAGE_SIZE) { | |
2227 | ret = -EINVAL; | |
2228 | goto free_inherit; | |
2229 | } | |
2230 | ||
2231 | nums = inherit->num_qgroups + 2 * inherit->num_ref_copies + | |
2232 | 2 * inherit->num_excl_copies; | |
2233 | if (vol_args->size != struct_size(inherit, qgroups, nums)) { | |
2234 | ret = -EINVAL; | |
2235 | goto free_inherit; | |
2236 | } | |
6f72c7e2 | 2237 | } |
fa0d2b9b | 2238 | |
4d4340c9 CB |
2239 | ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file), |
2240 | vol_args->name, vol_args->fd, subvol, | |
2241 | readonly, inherit); | |
c47ca32d DC |
2242 | if (ret) |
2243 | goto free_inherit; | |
c47ca32d | 2244 | free_inherit: |
6f72c7e2 | 2245 | kfree(inherit); |
c47ca32d DC |
2246 | free_args: |
2247 | kfree(vol_args); | |
f46b5a66 CH |
2248 | return ret; |
2249 | } | |
2250 | ||
9ad12305 | 2251 | static noinline int btrfs_ioctl_subvol_getflags(struct inode *inode, |
0caa102d LZ |
2252 | void __user *arg) |
2253 | { | |
0b246afa | 2254 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
0caa102d LZ |
2255 | struct btrfs_root *root = BTRFS_I(inode)->root; |
2256 | int ret = 0; | |
2257 | u64 flags = 0; | |
2258 | ||
4a0cc7ca | 2259 | if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) |
0caa102d LZ |
2260 | return -EINVAL; |
2261 | ||
0b246afa | 2262 | down_read(&fs_info->subvol_sem); |
0caa102d LZ |
2263 | if (btrfs_root_readonly(root)) |
2264 | flags |= BTRFS_SUBVOL_RDONLY; | |
0b246afa | 2265 | up_read(&fs_info->subvol_sem); |
0caa102d LZ |
2266 | |
2267 | if (copy_to_user(arg, &flags, sizeof(flags))) | |
2268 | ret = -EFAULT; | |
2269 | ||
2270 | return ret; | |
2271 | } | |
2272 | ||
2273 | static noinline int btrfs_ioctl_subvol_setflags(struct file *file, | |
2274 | void __user *arg) | |
2275 | { | |
496ad9aa | 2276 | struct inode *inode = file_inode(file); |
0b246afa | 2277 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
0caa102d LZ |
2278 | struct btrfs_root *root = BTRFS_I(inode)->root; |
2279 | struct btrfs_trans_handle *trans; | |
2280 | u64 root_flags; | |
2281 | u64 flags; | |
2282 | int ret = 0; | |
2283 | ||
39e1674f | 2284 | if (!inode_owner_or_capable(file_mnt_user_ns(file), inode)) |
bd60ea0f DS |
2285 | return -EPERM; |
2286 | ||
b9ca0664 LB |
2287 | ret = mnt_want_write_file(file); |
2288 | if (ret) | |
2289 | goto out; | |
0caa102d | 2290 | |
4a0cc7ca | 2291 | if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { |
b9ca0664 LB |
2292 | ret = -EINVAL; |
2293 | goto out_drop_write; | |
2294 | } | |
0caa102d | 2295 | |
b9ca0664 LB |
2296 | if (copy_from_user(&flags, arg, sizeof(flags))) { |
2297 | ret = -EFAULT; | |
2298 | goto out_drop_write; | |
2299 | } | |
0caa102d | 2300 | |
b9ca0664 LB |
2301 | if (flags & ~BTRFS_SUBVOL_RDONLY) { |
2302 | ret = -EOPNOTSUPP; | |
2303 | goto out_drop_write; | |
2304 | } | |
0caa102d | 2305 | |
0b246afa | 2306 | down_write(&fs_info->subvol_sem); |
0caa102d LZ |
2307 | |
2308 | /* nothing to do */ | |
2309 | if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) | |
b9ca0664 | 2310 | goto out_drop_sem; |
0caa102d LZ |
2311 | |
2312 | root_flags = btrfs_root_flags(&root->root_item); | |
2c686537 | 2313 | if (flags & BTRFS_SUBVOL_RDONLY) { |
0caa102d LZ |
2314 | btrfs_set_root_flags(&root->root_item, |
2315 | root_flags | BTRFS_ROOT_SUBVOL_RDONLY); | |
2c686537 DS |
2316 | } else { |
2317 | /* | |
2318 | * Block RO -> RW transition if this subvolume is involved in | |
2319 | * send | |
2320 | */ | |
2321 | spin_lock(&root->root_item_lock); | |
2322 | if (root->send_in_progress == 0) { | |
2323 | btrfs_set_root_flags(&root->root_item, | |
0caa102d | 2324 | root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); |
2c686537 DS |
2325 | spin_unlock(&root->root_item_lock); |
2326 | } else { | |
2327 | spin_unlock(&root->root_item_lock); | |
0b246afa JM |
2328 | btrfs_warn(fs_info, |
2329 | "Attempt to set subvolume %llu read-write during send", | |
2330 | root->root_key.objectid); | |
2c686537 DS |
2331 | ret = -EPERM; |
2332 | goto out_drop_sem; | |
2333 | } | |
2334 | } | |
0caa102d LZ |
2335 | |
2336 | trans = btrfs_start_transaction(root, 1); | |
2337 | if (IS_ERR(trans)) { | |
2338 | ret = PTR_ERR(trans); | |
2339 | goto out_reset; | |
2340 | } | |
2341 | ||
0b246afa | 2342 | ret = btrfs_update_root(trans, fs_info->tree_root, |
0caa102d | 2343 | &root->root_key, &root->root_item); |
9417ebc8 NB |
2344 | if (ret < 0) { |
2345 | btrfs_end_transaction(trans); | |
2346 | goto out_reset; | |
2347 | } | |
2348 | ||
2349 | ret = btrfs_commit_transaction(trans); | |
0caa102d | 2350 | |
0caa102d LZ |
2351 | out_reset: |
2352 | if (ret) | |
2353 | btrfs_set_root_flags(&root->root_item, root_flags); | |
b9ca0664 | 2354 | out_drop_sem: |
0b246afa | 2355 | up_write(&fs_info->subvol_sem); |
b9ca0664 LB |
2356 | out_drop_write: |
2357 | mnt_drop_write_file(file); | |
2358 | out: | |
0caa102d LZ |
2359 | return ret; |
2360 | } | |
2361 | ||
ac8e9819 CM |
2362 | static noinline int key_in_sk(struct btrfs_key *key, |
2363 | struct btrfs_ioctl_search_key *sk) | |
2364 | { | |
abc6e134 CM |
2365 | struct btrfs_key test; |
2366 | int ret; | |
2367 | ||
2368 | test.objectid = sk->min_objectid; | |
2369 | test.type = sk->min_type; | |
2370 | test.offset = sk->min_offset; | |
2371 | ||
2372 | ret = btrfs_comp_cpu_keys(key, &test); | |
2373 | if (ret < 0) | |
ac8e9819 | 2374 | return 0; |
abc6e134 CM |
2375 | |
2376 | test.objectid = sk->max_objectid; | |
2377 | test.type = sk->max_type; | |
2378 | test.offset = sk->max_offset; | |
2379 | ||
2380 | ret = btrfs_comp_cpu_keys(key, &test); | |
2381 | if (ret > 0) | |
ac8e9819 CM |
2382 | return 0; |
2383 | return 1; | |
2384 | } | |
2385 | ||
df397565 | 2386 | static noinline int copy_to_sk(struct btrfs_path *path, |
ac8e9819 CM |
2387 | struct btrfs_key *key, |
2388 | struct btrfs_ioctl_search_key *sk, | |
9b6e817d | 2389 | size_t *buf_size, |
ba346b35 | 2390 | char __user *ubuf, |
ac8e9819 CM |
2391 | unsigned long *sk_offset, |
2392 | int *num_found) | |
2393 | { | |
2394 | u64 found_transid; | |
2395 | struct extent_buffer *leaf; | |
2396 | struct btrfs_ioctl_search_header sh; | |
dd81d459 | 2397 | struct btrfs_key test; |
ac8e9819 CM |
2398 | unsigned long item_off; |
2399 | unsigned long item_len; | |
2400 | int nritems; | |
2401 | int i; | |
2402 | int slot; | |
ac8e9819 CM |
2403 | int ret = 0; |
2404 | ||
2405 | leaf = path->nodes[0]; | |
2406 | slot = path->slots[0]; | |
2407 | nritems = btrfs_header_nritems(leaf); | |
2408 | ||
2409 | if (btrfs_header_generation(leaf) > sk->max_transid) { | |
2410 | i = nritems; | |
2411 | goto advance_key; | |
2412 | } | |
2413 | found_transid = btrfs_header_generation(leaf); | |
2414 | ||
2415 | for (i = slot; i < nritems; i++) { | |
2416 | item_off = btrfs_item_ptr_offset(leaf, i); | |
3212fa14 | 2417 | item_len = btrfs_item_size(leaf, i); |
ac8e9819 | 2418 | |
03b71c6c GP |
2419 | btrfs_item_key_to_cpu(leaf, key, i); |
2420 | if (!key_in_sk(key, sk)) | |
2421 | continue; | |
2422 | ||
9b6e817d | 2423 | if (sizeof(sh) + item_len > *buf_size) { |
8f5f6178 GH |
2424 | if (*num_found) { |
2425 | ret = 1; | |
2426 | goto out; | |
2427 | } | |
2428 | ||
2429 | /* | |
2430 | * return one empty item back for v1, which does not | |
2431 | * handle -EOVERFLOW | |
2432 | */ | |
2433 | ||
9b6e817d | 2434 | *buf_size = sizeof(sh) + item_len; |
ac8e9819 | 2435 | item_len = 0; |
8f5f6178 GH |
2436 | ret = -EOVERFLOW; |
2437 | } | |
ac8e9819 | 2438 | |
9b6e817d | 2439 | if (sizeof(sh) + item_len + *sk_offset > *buf_size) { |
ac8e9819 | 2440 | ret = 1; |
25c9bc2e | 2441 | goto out; |
ac8e9819 CM |
2442 | } |
2443 | ||
ac8e9819 CM |
2444 | sh.objectid = key->objectid; |
2445 | sh.offset = key->offset; | |
2446 | sh.type = key->type; | |
2447 | sh.len = item_len; | |
2448 | sh.transid = found_transid; | |
2449 | ||
a48b73ec JB |
2450 | /* |
2451 | * Copy search result header. If we fault then loop again so we | |
2452 | * can fault in the pages and -EFAULT there if there's a | |
2453 | * problem. Otherwise we'll fault and then copy the buffer in | |
2454 | * properly this next time through | |
2455 | */ | |
2456 | if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) { | |
2457 | ret = 0; | |
ba346b35 GH |
2458 | goto out; |
2459 | } | |
2460 | ||
ac8e9819 CM |
2461 | *sk_offset += sizeof(sh); |
2462 | ||
2463 | if (item_len) { | |
ba346b35 | 2464 | char __user *up = ubuf + *sk_offset; |
a48b73ec JB |
2465 | /* |
2466 | * Copy the item, same behavior as above, but reset the | |
2467 | * * sk_offset so we copy the full thing again. | |
2468 | */ | |
2469 | if (read_extent_buffer_to_user_nofault(leaf, up, | |
2470 | item_off, item_len)) { | |
2471 | ret = 0; | |
2472 | *sk_offset -= sizeof(sh); | |
ba346b35 GH |
2473 | goto out; |
2474 | } | |
2475 | ||
ac8e9819 | 2476 | *sk_offset += item_len; |
ac8e9819 | 2477 | } |
e2156867 | 2478 | (*num_found)++; |
ac8e9819 | 2479 | |
8f5f6178 GH |
2480 | if (ret) /* -EOVERFLOW from above */ |
2481 | goto out; | |
2482 | ||
25c9bc2e GH |
2483 | if (*num_found >= sk->nr_items) { |
2484 | ret = 1; | |
2485 | goto out; | |
2486 | } | |
ac8e9819 CM |
2487 | } |
2488 | advance_key: | |
abc6e134 | 2489 | ret = 0; |
dd81d459 NA |
2490 | test.objectid = sk->max_objectid; |
2491 | test.type = sk->max_type; | |
2492 | test.offset = sk->max_offset; | |
2493 | if (btrfs_comp_cpu_keys(key, &test) >= 0) | |
2494 | ret = 1; | |
2495 | else if (key->offset < (u64)-1) | |
ac8e9819 | 2496 | key->offset++; |
dd81d459 | 2497 | else if (key->type < (u8)-1) { |
abc6e134 | 2498 | key->offset = 0; |
ac8e9819 | 2499 | key->type++; |
dd81d459 | 2500 | } else if (key->objectid < (u64)-1) { |
abc6e134 CM |
2501 | key->offset = 0; |
2502 | key->type = 0; | |
ac8e9819 | 2503 | key->objectid++; |
abc6e134 CM |
2504 | } else |
2505 | ret = 1; | |
25c9bc2e | 2506 | out: |
ba346b35 GH |
2507 | /* |
2508 | * 0: all items from this leaf copied, continue with next | |
2509 | * 1: * more items can be copied, but unused buffer is too small | |
2510 | * * all items were found | |
2511 | * Either way, it will stops the loop which iterates to the next | |
2512 | * leaf | |
2513 | * -EOVERFLOW: item was to large for buffer | |
2514 | * -EFAULT: could not copy extent buffer back to userspace | |
2515 | */ | |
ac8e9819 CM |
2516 | return ret; |
2517 | } | |
2518 | ||
2519 | static noinline int search_ioctl(struct inode *inode, | |
12544442 | 2520 | struct btrfs_ioctl_search_key *sk, |
9b6e817d | 2521 | size_t *buf_size, |
ba346b35 | 2522 | char __user *ubuf) |
ac8e9819 | 2523 | { |
0b246afa | 2524 | struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); |
ac8e9819 CM |
2525 | struct btrfs_root *root; |
2526 | struct btrfs_key key; | |
ac8e9819 | 2527 | struct btrfs_path *path; |
ac8e9819 CM |
2528 | int ret; |
2529 | int num_found = 0; | |
2530 | unsigned long sk_offset = 0; | |
2531 | ||
9b6e817d GH |
2532 | if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { |
2533 | *buf_size = sizeof(struct btrfs_ioctl_search_header); | |
12544442 | 2534 | return -EOVERFLOW; |
9b6e817d | 2535 | } |
12544442 | 2536 | |
ac8e9819 CM |
2537 | path = btrfs_alloc_path(); |
2538 | if (!path) | |
2539 | return -ENOMEM; | |
2540 | ||
2541 | if (sk->tree_id == 0) { | |
2542 | /* search the root of the inode that was passed */ | |
00246528 | 2543 | root = btrfs_grab_root(BTRFS_I(inode)->root); |
ac8e9819 | 2544 | } else { |
56e9357a | 2545 | root = btrfs_get_fs_root(info, sk->tree_id, true); |
ac8e9819 | 2546 | if (IS_ERR(root)) { |
ac8e9819 | 2547 | btrfs_free_path(path); |
ad1e3d56 | 2548 | return PTR_ERR(root); |
ac8e9819 CM |
2549 | } |
2550 | } | |
2551 | ||
2552 | key.objectid = sk->min_objectid; | |
2553 | key.type = sk->min_type; | |
2554 | key.offset = sk->min_offset; | |
2555 | ||
67871254 | 2556 | while (1) { |
bb523b40 AG |
2557 | ret = -EFAULT; |
2558 | if (fault_in_writeable(ubuf + sk_offset, *buf_size - sk_offset)) | |
a48b73ec JB |
2559 | break; |
2560 | ||
6174d3cb | 2561 | ret = btrfs_search_forward(root, &key, path, sk->min_transid); |
ac8e9819 CM |
2562 | if (ret != 0) { |
2563 | if (ret > 0) | |
2564 | ret = 0; | |
2565 | goto err; | |
2566 | } | |
df397565 | 2567 | ret = copy_to_sk(path, &key, sk, buf_size, ubuf, |
ac8e9819 | 2568 | &sk_offset, &num_found); |
b3b4aa74 | 2569 | btrfs_release_path(path); |
25c9bc2e | 2570 | if (ret) |
ac8e9819 CM |
2571 | break; |
2572 | ||
2573 | } | |
8f5f6178 GH |
2574 | if (ret > 0) |
2575 | ret = 0; | |
ac8e9819 CM |
2576 | err: |
2577 | sk->nr_items = num_found; | |
00246528 | 2578 | btrfs_put_root(root); |
ac8e9819 CM |
2579 | btrfs_free_path(path); |
2580 | return ret; | |
2581 | } | |
2582 | ||
9ad12305 SK |
2583 | static noinline int btrfs_ioctl_tree_search(struct inode *inode, |
2584 | void __user *argp) | |
ac8e9819 | 2585 | { |
ba346b35 GH |
2586 | struct btrfs_ioctl_search_args __user *uargs; |
2587 | struct btrfs_ioctl_search_key sk; | |
9b6e817d GH |
2588 | int ret; |
2589 | size_t buf_size; | |
ac8e9819 CM |
2590 | |
2591 | if (!capable(CAP_SYS_ADMIN)) | |
2592 | return -EPERM; | |
2593 | ||
ba346b35 GH |
2594 | uargs = (struct btrfs_ioctl_search_args __user *)argp; |
2595 | ||
2596 | if (copy_from_user(&sk, &uargs->key, sizeof(sk))) | |
2597 | return -EFAULT; | |
ac8e9819 | 2598 | |
ba346b35 | 2599 | buf_size = sizeof(uargs->buf); |
ac8e9819 | 2600 | |
ba346b35 | 2601 | ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); |
8f5f6178 GH |
2602 | |
2603 | /* | |
2604 | * In the origin implementation an overflow is handled by returning a | |
2605 | * search header with a len of zero, so reset ret. | |
2606 | */ | |
2607 | if (ret == -EOVERFLOW) | |
2608 | ret = 0; | |
2609 | ||
ba346b35 | 2610 | if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) |
ac8e9819 | 2611 | ret = -EFAULT; |
ac8e9819 CM |
2612 | return ret; |
2613 | } | |
2614 | ||
9ad12305 | 2615 | static noinline int btrfs_ioctl_tree_search_v2(struct inode *inode, |
cc68a8a5 GH |
2616 | void __user *argp) |
2617 | { | |
2618 | struct btrfs_ioctl_search_args_v2 __user *uarg; | |
2619 | struct btrfs_ioctl_search_args_v2 args; | |
cc68a8a5 GH |
2620 | int ret; |
2621 | size_t buf_size; | |
ee22184b | 2622 | const size_t buf_limit = SZ_16M; |
cc68a8a5 GH |
2623 | |
2624 | if (!capable(CAP_SYS_ADMIN)) | |
2625 | return -EPERM; | |
2626 | ||
2627 | /* copy search header and buffer size */ | |
2628 | uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; | |
2629 | if (copy_from_user(&args, uarg, sizeof(args))) | |
2630 | return -EFAULT; | |
2631 | ||
2632 | buf_size = args.buf_size; | |
2633 | ||
cc68a8a5 GH |
2634 | /* limit result size to 16MB */ |
2635 | if (buf_size > buf_limit) | |
2636 | buf_size = buf_limit; | |
2637 | ||
cc68a8a5 | 2638 | ret = search_ioctl(inode, &args.key, &buf_size, |
718dc5fa | 2639 | (char __user *)(&uarg->buf[0])); |
cc68a8a5 GH |
2640 | if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) |
2641 | ret = -EFAULT; | |
2642 | else if (ret == -EOVERFLOW && | |
2643 | copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) | |
2644 | ret = -EFAULT; | |
2645 | ||
ac8e9819 CM |
2646 | return ret; |
2647 | } | |
2648 | ||
98d377a0 | 2649 | /* |
ac8e9819 CM |
2650 | * Search INODE_REFs to identify path name of 'dirid' directory |
2651 | * in a 'tree_id' tree. and sets path name to 'name'. | |
2652 | */ | |
98d377a0 TH |
2653 | static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, |
2654 | u64 tree_id, u64 dirid, char *name) | |
2655 | { | |
2656 | struct btrfs_root *root; | |
2657 | struct btrfs_key key; | |
ac8e9819 | 2658 | char *ptr; |
98d377a0 TH |
2659 | int ret = -1; |
2660 | int slot; | |
2661 | int len; | |
2662 | int total_len = 0; | |
2663 | struct btrfs_inode_ref *iref; | |
2664 | struct extent_buffer *l; | |
2665 | struct btrfs_path *path; | |
2666 | ||
2667 | if (dirid == BTRFS_FIRST_FREE_OBJECTID) { | |
2668 | name[0]='\0'; | |
2669 | return 0; | |
2670 | } | |
2671 | ||
2672 | path = btrfs_alloc_path(); | |
2673 | if (!path) | |
2674 | return -ENOMEM; | |
2675 | ||
c8bcbfbd | 2676 | ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; |
98d377a0 | 2677 | |
56e9357a | 2678 | root = btrfs_get_fs_root(info, tree_id, true); |
98d377a0 | 2679 | if (IS_ERR(root)) { |
ad1e3d56 | 2680 | ret = PTR_ERR(root); |
88234012 JB |
2681 | root = NULL; |
2682 | goto out; | |
2683 | } | |
98d377a0 TH |
2684 | |
2685 | key.objectid = dirid; | |
2686 | key.type = BTRFS_INODE_REF_KEY; | |
8ad6fcab | 2687 | key.offset = (u64)-1; |
98d377a0 | 2688 | |
67871254 | 2689 | while (1) { |
0ff40a91 | 2690 | ret = btrfs_search_backwards(root, &key, path); |
98d377a0 TH |
2691 | if (ret < 0) |
2692 | goto out; | |
18674c6c | 2693 | else if (ret > 0) { |
0ff40a91 MPS |
2694 | ret = -ENOENT; |
2695 | goto out; | |
18674c6c | 2696 | } |
98d377a0 TH |
2697 | |
2698 | l = path->nodes[0]; | |
2699 | slot = path->slots[0]; | |
98d377a0 | 2700 | |
98d377a0 TH |
2701 | iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); |
2702 | len = btrfs_inode_ref_name_len(l, iref); | |
2703 | ptr -= len + 1; | |
2704 | total_len += len + 1; | |
a696cf35 FDBM |
2705 | if (ptr < name) { |
2706 | ret = -ENAMETOOLONG; | |
98d377a0 | 2707 | goto out; |
a696cf35 | 2708 | } |
98d377a0 TH |
2709 | |
2710 | *(ptr + len) = '/'; | |
67871254 | 2711 | read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); |
98d377a0 TH |
2712 | |
2713 | if (key.offset == BTRFS_FIRST_FREE_OBJECTID) | |
2714 | break; | |
2715 | ||
b3b4aa74 | 2716 | btrfs_release_path(path); |
98d377a0 | 2717 | key.objectid = key.offset; |
8ad6fcab | 2718 | key.offset = (u64)-1; |
98d377a0 | 2719 | dirid = key.objectid; |
98d377a0 | 2720 | } |
77906a50 | 2721 | memmove(name, ptr, total_len); |
67871254 | 2722 | name[total_len] = '\0'; |
98d377a0 TH |
2723 | ret = 0; |
2724 | out: | |
00246528 | 2725 | btrfs_put_root(root); |
98d377a0 | 2726 | btrfs_free_path(path); |
ac8e9819 CM |
2727 | return ret; |
2728 | } | |
2729 | ||
6623d9a0 CB |
2730 | static int btrfs_search_path_in_tree_user(struct user_namespace *mnt_userns, |
2731 | struct inode *inode, | |
23d0b79d TM |
2732 | struct btrfs_ioctl_ino_lookup_user_args *args) |
2733 | { | |
2734 | struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; | |
2735 | struct super_block *sb = inode->i_sb; | |
2736 | struct btrfs_key upper_limit = BTRFS_I(inode)->location; | |
2737 | u64 treeid = BTRFS_I(inode)->root->root_key.objectid; | |
2738 | u64 dirid = args->dirid; | |
2739 | unsigned long item_off; | |
2740 | unsigned long item_len; | |
2741 | struct btrfs_inode_ref *iref; | |
2742 | struct btrfs_root_ref *rref; | |
b8a49ae1 | 2743 | struct btrfs_root *root = NULL; |
23d0b79d TM |
2744 | struct btrfs_path *path; |
2745 | struct btrfs_key key, key2; | |
2746 | struct extent_buffer *leaf; | |
2747 | struct inode *temp_inode; | |
2748 | char *ptr; | |
2749 | int slot; | |
2750 | int len; | |
2751 | int total_len = 0; | |
2752 | int ret; | |
2753 | ||
2754 | path = btrfs_alloc_path(); | |
2755 | if (!path) | |
2756 | return -ENOMEM; | |
2757 | ||
2758 | /* | |
2759 | * If the bottom subvolume does not exist directly under upper_limit, | |
2760 | * construct the path in from the bottom up. | |
2761 | */ | |
2762 | if (dirid != upper_limit.objectid) { | |
2763 | ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1]; | |
2764 | ||
56e9357a | 2765 | root = btrfs_get_fs_root(fs_info, treeid, true); |
23d0b79d TM |
2766 | if (IS_ERR(root)) { |
2767 | ret = PTR_ERR(root); | |
2768 | goto out; | |
2769 | } | |
2770 | ||
2771 | key.objectid = dirid; | |
2772 | key.type = BTRFS_INODE_REF_KEY; | |
2773 | key.offset = (u64)-1; | |
2774 | while (1) { | |
0ff40a91 MPS |
2775 | ret = btrfs_search_backwards(root, &key, path); |
2776 | if (ret < 0) | |
2777 | goto out_put; | |
2778 | else if (ret > 0) { | |
2779 | ret = -ENOENT; | |
b8a49ae1 | 2780 | goto out_put; |
23d0b79d TM |
2781 | } |
2782 | ||
2783 | leaf = path->nodes[0]; | |
2784 | slot = path->slots[0]; | |
23d0b79d TM |
2785 | |
2786 | iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref); | |
2787 | len = btrfs_inode_ref_name_len(leaf, iref); | |
2788 | ptr -= len + 1; | |
2789 | total_len += len + 1; | |
2790 | if (ptr < args->path) { | |
2791 | ret = -ENAMETOOLONG; | |
b8a49ae1 | 2792 | goto out_put; |
23d0b79d TM |
2793 | } |
2794 | ||
2795 | *(ptr + len) = '/'; | |
2796 | read_extent_buffer(leaf, ptr, | |
2797 | (unsigned long)(iref + 1), len); | |
2798 | ||
2799 | /* Check the read+exec permission of this directory */ | |
2800 | ret = btrfs_previous_item(root, path, dirid, | |
2801 | BTRFS_INODE_ITEM_KEY); | |
2802 | if (ret < 0) { | |
b8a49ae1 | 2803 | goto out_put; |
23d0b79d TM |
2804 | } else if (ret > 0) { |
2805 | ret = -ENOENT; | |
b8a49ae1 | 2806 | goto out_put; |
23d0b79d TM |
2807 | } |
2808 | ||
2809 | leaf = path->nodes[0]; | |
2810 | slot = path->slots[0]; | |
2811 | btrfs_item_key_to_cpu(leaf, &key2, slot); | |
2812 | if (key2.objectid != dirid) { | |
2813 | ret = -ENOENT; | |
b8a49ae1 | 2814 | goto out_put; |
23d0b79d TM |
2815 | } |
2816 | ||
0202e83f | 2817 | temp_inode = btrfs_iget(sb, key2.objectid, root); |
3ca57bd6 MT |
2818 | if (IS_ERR(temp_inode)) { |
2819 | ret = PTR_ERR(temp_inode); | |
b8a49ae1 | 2820 | goto out_put; |
3ca57bd6 | 2821 | } |
6623d9a0 | 2822 | ret = inode_permission(mnt_userns, temp_inode, |
47291baa | 2823 | MAY_READ | MAY_EXEC); |
23d0b79d TM |
2824 | iput(temp_inode); |
2825 | if (ret) { | |
2826 | ret = -EACCES; | |
b8a49ae1 | 2827 | goto out_put; |
23d0b79d TM |
2828 | } |
2829 | ||
2830 | if (key.offset == upper_limit.objectid) | |
2831 | break; | |
2832 | if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) { | |
2833 | ret = -EACCES; | |
b8a49ae1 | 2834 | goto out_put; |
23d0b79d TM |
2835 | } |
2836 | ||
2837 | btrfs_release_path(path); | |
2838 | key.objectid = key.offset; | |
2839 | key.offset = (u64)-1; | |
2840 | dirid = key.objectid; | |
2841 | } | |
2842 | ||
2843 | memmove(args->path, ptr, total_len); | |
2844 | args->path[total_len] = '\0'; | |
00246528 | 2845 | btrfs_put_root(root); |
b8a49ae1 | 2846 | root = NULL; |
23d0b79d TM |
2847 | btrfs_release_path(path); |
2848 | } | |
2849 | ||
2850 | /* Get the bottom subvolume's name from ROOT_REF */ | |
23d0b79d TM |
2851 | key.objectid = treeid; |
2852 | key.type = BTRFS_ROOT_REF_KEY; | |
2853 | key.offset = args->treeid; | |
b8a49ae1 | 2854 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); |
23d0b79d TM |
2855 | if (ret < 0) { |
2856 | goto out; | |
2857 | } else if (ret > 0) { | |
2858 | ret = -ENOENT; | |
2859 | goto out; | |
2860 | } | |
2861 | ||
2862 | leaf = path->nodes[0]; | |
2863 | slot = path->slots[0]; | |
2864 | btrfs_item_key_to_cpu(leaf, &key, slot); | |
2865 | ||
2866 | item_off = btrfs_item_ptr_offset(leaf, slot); | |
3212fa14 | 2867 | item_len = btrfs_item_size(leaf, slot); |
23d0b79d TM |
2868 | /* Check if dirid in ROOT_REF corresponds to passed dirid */ |
2869 | rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); | |
2870 | if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) { | |
2871 | ret = -EINVAL; | |
2872 | goto out; | |
2873 | } | |
2874 | ||
2875 | /* Copy subvolume's name */ | |
2876 | item_off += sizeof(struct btrfs_root_ref); | |
2877 | item_len -= sizeof(struct btrfs_root_ref); | |
2878 | read_extent_buffer(leaf, args->name, item_off, item_len); | |
2879 | args->name[item_len] = 0; | |
2880 | ||
b8a49ae1 | 2881 | out_put: |
00246528 | 2882 | btrfs_put_root(root); |
23d0b79d TM |
2883 | out: |
2884 | btrfs_free_path(path); | |
2885 | return ret; | |
2886 | } | |
2887 | ||
dc408ccd | 2888 | static noinline int btrfs_ioctl_ino_lookup(struct btrfs_root *root, |
ac8e9819 CM |
2889 | void __user *argp) |
2890 | { | |
bece2e82 | 2891 | struct btrfs_ioctl_ino_lookup_args *args; |
01b810b8 | 2892 | int ret = 0; |
ac8e9819 | 2893 | |
2354d08f JL |
2894 | args = memdup_user(argp, sizeof(*args)); |
2895 | if (IS_ERR(args)) | |
2896 | return PTR_ERR(args); | |
c2b96929 | 2897 | |
01b810b8 DS |
2898 | /* |
2899 | * Unprivileged query to obtain the containing subvolume root id. The | |
2900 | * path is reset so it's consistent with btrfs_search_path_in_tree. | |
2901 | */ | |
1b53ac4d | 2902 | if (args->treeid == 0) |
dc408ccd | 2903 | args->treeid = root->root_key.objectid; |
1b53ac4d | 2904 | |
01b810b8 DS |
2905 | if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { |
2906 | args->name[0] = 0; | |
2907 | goto out; | |
2908 | } | |
2909 | ||
2910 | if (!capable(CAP_SYS_ADMIN)) { | |
2911 | ret = -EPERM; | |
2912 | goto out; | |
2913 | } | |
2914 | ||
dc408ccd | 2915 | ret = btrfs_search_path_in_tree(root->fs_info, |
ac8e9819 CM |
2916 | args->treeid, args->objectid, |
2917 | args->name); | |
2918 | ||
01b810b8 | 2919 | out: |
ac8e9819 CM |
2920 | if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) |
2921 | ret = -EFAULT; | |
2922 | ||
2923 | kfree(args); | |
98d377a0 TH |
2924 | return ret; |
2925 | } | |
2926 | ||
23d0b79d TM |
2927 | /* |
2928 | * Version of ino_lookup ioctl (unprivileged) | |
2929 | * | |
2930 | * The main differences from ino_lookup ioctl are: | |
2931 | * | |
2932 | * 1. Read + Exec permission will be checked using inode_permission() during | |
2933 | * path construction. -EACCES will be returned in case of failure. | |
2934 | * 2. Path construction will be stopped at the inode number which corresponds | |
2935 | * to the fd with which this ioctl is called. If constructed path does not | |
2936 | * exist under fd's inode, -EACCES will be returned. | |
2937 | * 3. The name of bottom subvolume is also searched and filled. | |
2938 | */ | |
2939 | static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp) | |
2940 | { | |
2941 | struct btrfs_ioctl_ino_lookup_user_args *args; | |
2942 | struct inode *inode; | |
2943 | int ret; | |
2944 | ||
2945 | args = memdup_user(argp, sizeof(*args)); | |
2946 | if (IS_ERR(args)) | |
2947 | return PTR_ERR(args); | |
2948 | ||
2949 | inode = file_inode(file); | |
2950 | ||
2951 | if (args->dirid == BTRFS_FIRST_FREE_OBJECTID && | |
2952 | BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) { | |
2953 | /* | |
2954 | * The subvolume does not exist under fd with which this is | |
2955 | * called | |
2956 | */ | |
2957 | kfree(args); | |
2958 | return -EACCES; | |
2959 | } | |
2960 | ||
6623d9a0 | 2961 | ret = btrfs_search_path_in_tree_user(file_mnt_user_ns(file), inode, args); |
23d0b79d TM |
2962 | |
2963 | if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) | |
2964 | ret = -EFAULT; | |
2965 | ||
2966 | kfree(args); | |
2967 | return ret; | |
2968 | } | |
2969 | ||
b64ec075 | 2970 | /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */ |
9ad12305 | 2971 | static int btrfs_ioctl_get_subvol_info(struct inode *inode, void __user *argp) |
b64ec075 TM |
2972 | { |
2973 | struct btrfs_ioctl_get_subvol_info_args *subvol_info; | |
2974 | struct btrfs_fs_info *fs_info; | |
2975 | struct btrfs_root *root; | |
2976 | struct btrfs_path *path; | |
2977 | struct btrfs_key key; | |
2978 | struct btrfs_root_item *root_item; | |
2979 | struct btrfs_root_ref *rref; | |
2980 | struct extent_buffer *leaf; | |
2981 | unsigned long item_off; | |
2982 | unsigned long item_len; | |
b64ec075 TM |
2983 | int slot; |
2984 | int ret = 0; | |
2985 | ||
2986 | path = btrfs_alloc_path(); | |
2987 | if (!path) | |
2988 | return -ENOMEM; | |
2989 | ||
2990 | subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL); | |
2991 | if (!subvol_info) { | |
2992 | btrfs_free_path(path); | |
2993 | return -ENOMEM; | |
2994 | } | |
2995 | ||
b64ec075 TM |
2996 | fs_info = BTRFS_I(inode)->root->fs_info; |
2997 | ||
2998 | /* Get root_item of inode's subvolume */ | |
2999 | key.objectid = BTRFS_I(inode)->root->root_key.objectid; | |
56e9357a | 3000 | root = btrfs_get_fs_root(fs_info, key.objectid, true); |
b64ec075 TM |
3001 | if (IS_ERR(root)) { |
3002 | ret = PTR_ERR(root); | |
04734e84 JB |
3003 | goto out_free; |
3004 | } | |
b64ec075 TM |
3005 | root_item = &root->root_item; |
3006 | ||
3007 | subvol_info->treeid = key.objectid; | |
3008 | ||
3009 | subvol_info->generation = btrfs_root_generation(root_item); | |
3010 | subvol_info->flags = btrfs_root_flags(root_item); | |
3011 | ||
3012 | memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE); | |
3013 | memcpy(subvol_info->parent_uuid, root_item->parent_uuid, | |
3014 | BTRFS_UUID_SIZE); | |
3015 | memcpy(subvol_info->received_uuid, root_item->received_uuid, | |
3016 | BTRFS_UUID_SIZE); | |
3017 | ||
3018 | subvol_info->ctransid = btrfs_root_ctransid(root_item); | |
3019 | subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime); | |
3020 | subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime); | |
3021 | ||
3022 | subvol_info->otransid = btrfs_root_otransid(root_item); | |
3023 | subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime); | |
3024 | subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime); | |
3025 | ||
3026 | subvol_info->stransid = btrfs_root_stransid(root_item); | |
3027 | subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime); | |
3028 | subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime); | |
3029 | ||
3030 | subvol_info->rtransid = btrfs_root_rtransid(root_item); | |
3031 | subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime); | |
3032 | subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime); | |
3033 | ||
3034 | if (key.objectid != BTRFS_FS_TREE_OBJECTID) { | |
3035 | /* Search root tree for ROOT_BACKREF of this subvolume */ | |
b64ec075 TM |
3036 | key.type = BTRFS_ROOT_BACKREF_KEY; |
3037 | key.offset = 0; | |
04734e84 | 3038 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); |
b64ec075 TM |
3039 | if (ret < 0) { |
3040 | goto out; | |
3041 | } else if (path->slots[0] >= | |
3042 | btrfs_header_nritems(path->nodes[0])) { | |
04734e84 | 3043 | ret = btrfs_next_leaf(fs_info->tree_root, path); |
b64ec075 TM |
3044 | if (ret < 0) { |
3045 | goto out; | |
3046 | } else if (ret > 0) { | |
3047 | ret = -EUCLEAN; | |
3048 | goto out; | |
3049 | } | |
3050 | } | |
3051 | ||
3052 | leaf = path->nodes[0]; | |
3053 | slot = path->slots[0]; | |
3054 | btrfs_item_key_to_cpu(leaf, &key, slot); | |
3055 | if (key.objectid == subvol_info->treeid && | |
3056 | key.type == BTRFS_ROOT_BACKREF_KEY) { | |
3057 | subvol_info->parent_id = key.offset; | |
3058 | ||
3059 | rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); | |
3060 | subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref); | |
3061 | ||
3062 | item_off = btrfs_item_ptr_offset(leaf, slot) | |
3063 | + sizeof(struct btrfs_root_ref); | |
3212fa14 | 3064 | item_len = btrfs_item_size(leaf, slot) |
b64ec075 TM |
3065 | - sizeof(struct btrfs_root_ref); |
3066 | read_extent_buffer(leaf, subvol_info->name, | |
3067 | item_off, item_len); | |
3068 | } else { | |
3069 | ret = -ENOENT; | |
3070 | goto out; | |
3071 | } | |
3072 | } | |
3073 | ||
3074 | if (copy_to_user(argp, subvol_info, sizeof(*subvol_info))) | |
3075 | ret = -EFAULT; | |
3076 | ||
3077 | out: | |
00246528 | 3078 | btrfs_put_root(root); |
04734e84 | 3079 | out_free: |
b64ec075 | 3080 | btrfs_free_path(path); |
b091f7fe | 3081 | kfree(subvol_info); |
b64ec075 TM |
3082 | return ret; |
3083 | } | |
3084 | ||
42e4b520 TM |
3085 | /* |
3086 | * Return ROOT_REF information of the subvolume containing this inode | |
3087 | * except the subvolume name. | |
3088 | */ | |
9ad12305 | 3089 | static int btrfs_ioctl_get_subvol_rootref(struct btrfs_root *root, |
dc408ccd | 3090 | void __user *argp) |
42e4b520 TM |
3091 | { |
3092 | struct btrfs_ioctl_get_subvol_rootref_args *rootrefs; | |
3093 | struct btrfs_root_ref *rref; | |
42e4b520 TM |
3094 | struct btrfs_path *path; |
3095 | struct btrfs_key key; | |
3096 | struct extent_buffer *leaf; | |
42e4b520 TM |
3097 | u64 objectid; |
3098 | int slot; | |
3099 | int ret; | |
3100 | u8 found; | |
3101 | ||
3102 | path = btrfs_alloc_path(); | |
3103 | if (!path) | |
3104 | return -ENOMEM; | |
3105 | ||
3106 | rootrefs = memdup_user(argp, sizeof(*rootrefs)); | |
3107 | if (IS_ERR(rootrefs)) { | |
3108 | btrfs_free_path(path); | |
3109 | return PTR_ERR(rootrefs); | |
3110 | } | |
3111 | ||
9ad12305 | 3112 | objectid = root->root_key.objectid; |
42e4b520 TM |
3113 | key.objectid = objectid; |
3114 | key.type = BTRFS_ROOT_REF_KEY; | |
3115 | key.offset = rootrefs->min_treeid; | |
3116 | found = 0; | |
3117 | ||
9ad12305 | 3118 | root = root->fs_info->tree_root; |
42e4b520 TM |
3119 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
3120 | if (ret < 0) { | |
3121 | goto out; | |
3122 | } else if (path->slots[0] >= | |
3123 | btrfs_header_nritems(path->nodes[0])) { | |
3124 | ret = btrfs_next_leaf(root, path); | |
3125 | if (ret < 0) { | |
3126 | goto out; | |
3127 | } else if (ret > 0) { | |
3128 | ret = -EUCLEAN; | |
3129 | goto out; | |
3130 | } | |
3131 | } | |
3132 | while (1) { | |
3133 | leaf = path->nodes[0]; | |
3134 | slot = path->slots[0]; | |
3135 | ||
3136 | btrfs_item_key_to_cpu(leaf, &key, slot); | |
3137 | if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) { | |
3138 | ret = 0; | |
3139 | goto out; | |
3140 | } | |
3141 | ||
3142 | if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) { | |
3143 | ret = -EOVERFLOW; | |
3144 | goto out; | |
3145 | } | |
3146 | ||
3147 | rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); | |
3148 | rootrefs->rootref[found].treeid = key.offset; | |
3149 | rootrefs->rootref[found].dirid = | |
3150 | btrfs_root_ref_dirid(leaf, rref); | |
3151 | found++; | |
3152 | ||
3153 | ret = btrfs_next_item(root, path); | |
3154 | if (ret < 0) { | |
3155 | goto out; | |
3156 | } else if (ret > 0) { | |
3157 | ret = -EUCLEAN; | |
3158 | goto out; | |
3159 | } | |
3160 | } | |
3161 | ||
3162 | out: | |
3163 | if (!ret || ret == -EOVERFLOW) { | |
3164 | rootrefs->num_items = found; | |
3165 | /* update min_treeid for next search */ | |
3166 | if (found) | |
3167 | rootrefs->min_treeid = | |
3168 | rootrefs->rootref[found - 1].treeid + 1; | |
3169 | if (copy_to_user(argp, rootrefs, sizeof(*rootrefs))) | |
3170 | ret = -EFAULT; | |
3171 | } | |
3172 | ||
3173 | kfree(rootrefs); | |
3174 | btrfs_free_path(path); | |
3175 | ||
3176 | return ret; | |
3177 | } | |
3178 | ||
76dda93c | 3179 | static noinline int btrfs_ioctl_snap_destroy(struct file *file, |
949964c9 MPS |
3180 | void __user *arg, |
3181 | bool destroy_v2) | |
76dda93c | 3182 | { |
54563d41 | 3183 | struct dentry *parent = file->f_path.dentry; |
0b246afa | 3184 | struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); |
76dda93c | 3185 | struct dentry *dentry; |
2b0143b5 | 3186 | struct inode *dir = d_inode(parent); |
76dda93c YZ |
3187 | struct inode *inode; |
3188 | struct btrfs_root *root = BTRFS_I(dir)->root; | |
3189 | struct btrfs_root *dest = NULL; | |
949964c9 MPS |
3190 | struct btrfs_ioctl_vol_args *vol_args = NULL; |
3191 | struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL; | |
c4ed533b | 3192 | struct user_namespace *mnt_userns = file_mnt_user_ns(file); |
949964c9 MPS |
3193 | char *subvol_name, *subvol_name_ptr = NULL; |
3194 | int subvol_namelen; | |
76dda93c | 3195 | int err = 0; |
949964c9 | 3196 | bool destroy_parent = false; |
76dda93c | 3197 | |
813febdb JB |
3198 | /* We don't support snapshots with extent tree v2 yet. */ |
3199 | if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { | |
3200 | btrfs_err(fs_info, | |
3201 | "extent tree v2 doesn't support snapshot deletion yet"); | |
3202 | return -EOPNOTSUPP; | |
3203 | } | |
3204 | ||
949964c9 MPS |
3205 | if (destroy_v2) { |
3206 | vol_args2 = memdup_user(arg, sizeof(*vol_args2)); | |
3207 | if (IS_ERR(vol_args2)) | |
3208 | return PTR_ERR(vol_args2); | |
325c50e3 | 3209 | |
949964c9 MPS |
3210 | if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) { |
3211 | err = -EOPNOTSUPP; | |
3212 | goto out; | |
3213 | } | |
76dda93c | 3214 | |
949964c9 MPS |
3215 | /* |
3216 | * If SPEC_BY_ID is not set, we are looking for the subvolume by | |
3217 | * name, same as v1 currently does. | |
3218 | */ | |
3219 | if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) { | |
3220 | vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0; | |
3221 | subvol_name = vol_args2->name; | |
3222 | ||
3223 | err = mnt_want_write_file(file); | |
3224 | if (err) | |
3225 | goto out; | |
3226 | } else { | |
aabb34e7 | 3227 | struct inode *old_dir; |
c4ed533b | 3228 | |
949964c9 MPS |
3229 | if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) { |
3230 | err = -EINVAL; | |
3231 | goto out; | |
3232 | } | |
3233 | ||
3234 | err = mnt_want_write_file(file); | |
3235 | if (err) | |
3236 | goto out; | |
3237 | ||
3238 | dentry = btrfs_get_dentry(fs_info->sb, | |
3239 | BTRFS_FIRST_FREE_OBJECTID, | |
3240 | vol_args2->subvolid, 0, 0); | |
3241 | if (IS_ERR(dentry)) { | |
3242 | err = PTR_ERR(dentry); | |
3243 | goto out_drop_write; | |
3244 | } | |
3245 | ||
3246 | /* | |
3247 | * Change the default parent since the subvolume being | |
3248 | * deleted can be outside of the current mount point. | |
3249 | */ | |
3250 | parent = btrfs_get_parent(dentry); | |
3251 | ||
3252 | /* | |
3253 | * At this point dentry->d_name can point to '/' if the | |
3254 | * subvolume we want to destroy is outsite of the | |
3255 | * current mount point, so we need to release the | |
3256 | * current dentry and execute the lookup to return a new | |
3257 | * one with ->d_name pointing to the | |
3258 | * <mount point>/subvol_name. | |
3259 | */ | |
3260 | dput(dentry); | |
3261 | if (IS_ERR(parent)) { | |
3262 | err = PTR_ERR(parent); | |
3263 | goto out_drop_write; | |
3264 | } | |
aabb34e7 | 3265 | old_dir = dir; |
949964c9 MPS |
3266 | dir = d_inode(parent); |
3267 | ||
3268 | /* | |
3269 | * If v2 was used with SPEC_BY_ID, a new parent was | |
3270 | * allocated since the subvolume can be outside of the | |
3271 | * current mount point. Later on we need to release this | |
3272 | * new parent dentry. | |
3273 | */ | |
3274 | destroy_parent = true; | |
3275 | ||
aabb34e7 CB |
3276 | /* |
3277 | * On idmapped mounts, deletion via subvolid is | |
3278 | * restricted to subvolumes that are immediate | |
3279 | * ancestors of the inode referenced by the file | |
3280 | * descriptor in the ioctl. Otherwise the idmapping | |
3281 | * could potentially be abused to delete subvolumes | |
3282 | * anywhere in the filesystem the user wouldn't be able | |
3283 | * to delete without an idmapped mount. | |
3284 | */ | |
3285 | if (old_dir != dir && mnt_userns != &init_user_ns) { | |
3286 | err = -EOPNOTSUPP; | |
3287 | goto free_parent; | |
3288 | } | |
3289 | ||
949964c9 MPS |
3290 | subvol_name_ptr = btrfs_get_subvol_name_from_objectid( |
3291 | fs_info, vol_args2->subvolid); | |
3292 | if (IS_ERR(subvol_name_ptr)) { | |
3293 | err = PTR_ERR(subvol_name_ptr); | |
3294 | goto free_parent; | |
3295 | } | |
1a9fd417 | 3296 | /* subvol_name_ptr is already nul terminated */ |
949964c9 MPS |
3297 | subvol_name = (char *)kbasename(subvol_name_ptr); |
3298 | } | |
3299 | } else { | |
3300 | vol_args = memdup_user(arg, sizeof(*vol_args)); | |
3301 | if (IS_ERR(vol_args)) | |
3302 | return PTR_ERR(vol_args); | |
3303 | ||
3304 | vol_args->name[BTRFS_PATH_NAME_MAX] = 0; | |
3305 | subvol_name = vol_args->name; | |
3306 | ||
3307 | err = mnt_want_write_file(file); | |
3308 | if (err) | |
3309 | goto out; | |
76dda93c YZ |
3310 | } |
3311 | ||
949964c9 | 3312 | subvol_namelen = strlen(subvol_name); |
76dda93c | 3313 | |
949964c9 MPS |
3314 | if (strchr(subvol_name, '/') || |
3315 | strncmp(subvol_name, "..", subvol_namelen) == 0) { | |
3316 | err = -EINVAL; | |
3317 | goto free_subvol_name; | |
3318 | } | |
3319 | ||
3320 | if (!S_ISDIR(dir->i_mode)) { | |
3321 | err = -ENOTDIR; | |
3322 | goto free_subvol_name; | |
3323 | } | |
521e0546 | 3324 | |
00235411 AV |
3325 | err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); |
3326 | if (err == -EINTR) | |
949964c9 | 3327 | goto free_subvol_name; |
c4ed533b | 3328 | dentry = lookup_one(mnt_userns, subvol_name, parent, subvol_namelen); |
76dda93c YZ |
3329 | if (IS_ERR(dentry)) { |
3330 | err = PTR_ERR(dentry); | |
3331 | goto out_unlock_dir; | |
3332 | } | |
3333 | ||
2b0143b5 | 3334 | if (d_really_is_negative(dentry)) { |
76dda93c YZ |
3335 | err = -ENOENT; |
3336 | goto out_dput; | |
3337 | } | |
3338 | ||
2b0143b5 | 3339 | inode = d_inode(dentry); |
4260f7c7 | 3340 | dest = BTRFS_I(inode)->root; |
67871254 | 3341 | if (!capable(CAP_SYS_ADMIN)) { |
4260f7c7 SW |
3342 | /* |
3343 | * Regular user. Only allow this with a special mount | |
3344 | * option, when the user has write+exec access to the | |
3345 | * subvol root, and when rmdir(2) would have been | |
3346 | * allowed. | |
3347 | * | |
3348 | * Note that this is _not_ check that the subvol is | |
3349 | * empty or doesn't contain data that we wouldn't | |
3350 | * otherwise be able to delete. | |
3351 | * | |
3352 | * Users who want to delete empty subvols should try | |
3353 | * rmdir(2). | |
3354 | */ | |
3355 | err = -EPERM; | |
0b246afa | 3356 | if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) |
4260f7c7 SW |
3357 | goto out_dput; |
3358 | ||
3359 | /* | |
3360 | * Do not allow deletion if the parent dir is the same | |
3361 | * as the dir to be deleted. That means the ioctl | |
3362 | * must be called on the dentry referencing the root | |
3363 | * of the subvol, not a random directory contained | |
3364 | * within it. | |
3365 | */ | |
3366 | err = -EINVAL; | |
3367 | if (root == dest) | |
3368 | goto out_dput; | |
3369 | ||
c4ed533b | 3370 | err = inode_permission(mnt_userns, inode, MAY_WRITE | MAY_EXEC); |
4260f7c7 SW |
3371 | if (err) |
3372 | goto out_dput; | |
4260f7c7 SW |
3373 | } |
3374 | ||
5c39da5b | 3375 | /* check if subvolume may be deleted by a user */ |
c4ed533b | 3376 | err = btrfs_may_delete(mnt_userns, dir, dentry, 1); |
5c39da5b MX |
3377 | if (err) |
3378 | goto out_dput; | |
3379 | ||
4a0cc7ca | 3380 | if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { |
76dda93c YZ |
3381 | err = -EINVAL; |
3382 | goto out_dput; | |
3383 | } | |
3384 | ||
64708539 | 3385 | btrfs_inode_lock(inode, 0); |
f60a2364 | 3386 | err = btrfs_delete_subvolume(dir, dentry); |
64708539 | 3387 | btrfs_inode_unlock(inode, 0); |
a37d9a17 AG |
3388 | if (!err) |
3389 | d_delete_notify(dir, dentry); | |
fa6ac876 | 3390 | |
76dda93c YZ |
3391 | out_dput: |
3392 | dput(dentry); | |
3393 | out_unlock_dir: | |
64708539 | 3394 | btrfs_inode_unlock(dir, 0); |
949964c9 MPS |
3395 | free_subvol_name: |
3396 | kfree(subvol_name_ptr); | |
3397 | free_parent: | |
3398 | if (destroy_parent) | |
3399 | dput(parent); | |
00235411 | 3400 | out_drop_write: |
2a79f17e | 3401 | mnt_drop_write_file(file); |
76dda93c | 3402 | out: |
949964c9 | 3403 | kfree(vol_args2); |
76dda93c YZ |
3404 | kfree(vol_args); |
3405 | return err; | |
3406 | } | |
3407 | ||
1e701a32 | 3408 | static int btrfs_ioctl_defrag(struct file *file, void __user *argp) |
f46b5a66 | 3409 | { |
496ad9aa | 3410 | struct inode *inode = file_inode(file); |
f46b5a66 | 3411 | struct btrfs_root *root = BTRFS_I(inode)->root; |
c853a578 | 3412 | struct btrfs_ioctl_defrag_range_args range = {0}; |
c146afad YZ |
3413 | int ret; |
3414 | ||
25122d15 ID |
3415 | ret = mnt_want_write_file(file); |
3416 | if (ret) | |
3417 | return ret; | |
b83cc969 | 3418 | |
25122d15 ID |
3419 | if (btrfs_root_readonly(root)) { |
3420 | ret = -EROFS; | |
3421 | goto out; | |
5ac00add | 3422 | } |
f46b5a66 CH |
3423 | |
3424 | switch (inode->i_mode & S_IFMT) { | |
3425 | case S_IFDIR: | |
e441d54d CM |
3426 | if (!capable(CAP_SYS_ADMIN)) { |
3427 | ret = -EPERM; | |
3428 | goto out; | |
3429 | } | |
de78b51a | 3430 | ret = btrfs_defrag_root(root); |
f46b5a66 CH |
3431 | break; |
3432 | case S_IFREG: | |
616d374e AB |
3433 | /* |
3434 | * Note that this does not check the file descriptor for write | |
3435 | * access. This prevents defragmenting executables that are | |
3436 | * running and allows defrag on files open in read-only mode. | |
3437 | */ | |
3438 | if (!capable(CAP_SYS_ADMIN) && | |
47291baa | 3439 | inode_permission(&init_user_ns, inode, MAY_WRITE)) { |
616d374e | 3440 | ret = -EPERM; |
e441d54d CM |
3441 | goto out; |
3442 | } | |
1e701a32 | 3443 | |
1e701a32 | 3444 | if (argp) { |
c853a578 | 3445 | if (copy_from_user(&range, argp, sizeof(range))) { |
1e701a32 | 3446 | ret = -EFAULT; |
683be16e | 3447 | goto out; |
1e701a32 CM |
3448 | } |
3449 | /* compression requires us to start the IO */ | |
c853a578 GR |
3450 | if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { |
3451 | range.flags |= BTRFS_DEFRAG_RANGE_START_IO; | |
3452 | range.extent_thresh = (u32)-1; | |
1e701a32 CM |
3453 | } |
3454 | } else { | |
3455 | /* the rest are all set to zero by kzalloc */ | |
c853a578 | 3456 | range.len = (u64)-1; |
1e701a32 | 3457 | } |
1ccc2e8a | 3458 | ret = btrfs_defrag_file(file_inode(file), &file->f_ra, |
c853a578 | 3459 | &range, BTRFS_OLDEST_GENERATION, 0); |
4cb5300b CM |
3460 | if (ret > 0) |
3461 | ret = 0; | |
f46b5a66 | 3462 | break; |
8929ecfa YZ |
3463 | default: |
3464 | ret = -EINVAL; | |
f46b5a66 | 3465 | } |
e441d54d | 3466 | out: |
25122d15 | 3467 | mnt_drop_write_file(file); |
e441d54d | 3468 | return ret; |
f46b5a66 CH |
3469 | } |
3470 | ||
2ff7e61e | 3471 | static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) |
f46b5a66 CH |
3472 | { |
3473 | struct btrfs_ioctl_vol_args *vol_args; | |
a174c0a2 | 3474 | bool restore_op = false; |
f46b5a66 CH |
3475 | int ret; |
3476 | ||
e441d54d CM |
3477 | if (!capable(CAP_SYS_ADMIN)) |
3478 | return -EPERM; | |
3479 | ||
914a519b JB |
3480 | if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { |
3481 | btrfs_err(fs_info, "device add not supported on extent tree v2 yet"); | |
3482 | return -EINVAL; | |
3483 | } | |
3484 | ||
a174c0a2 NB |
3485 | if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) { |
3486 | if (!btrfs_exclop_start_try_lock(fs_info, BTRFS_EXCLOP_DEV_ADD)) | |
3487 | return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; | |
3488 | ||
3489 | /* | |
3490 | * We can do the device add because we have a paused balanced, | |
3491 | * change the exclusive op type and remember we should bring | |
3492 | * back the paused balance | |
3493 | */ | |
3494 | fs_info->exclusive_operation = BTRFS_EXCLOP_DEV_ADD; | |
3495 | btrfs_exclop_start_unlock(fs_info); | |
3496 | restore_op = true; | |
3497 | } | |
c9e9f97b | 3498 | |
dae7b665 | 3499 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
3500 | if (IS_ERR(vol_args)) { |
3501 | ret = PTR_ERR(vol_args); | |
3502 | goto out; | |
3503 | } | |
f46b5a66 | 3504 | |
5516e595 | 3505 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
2ff7e61e | 3506 | ret = btrfs_init_new_device(fs_info, vol_args->name); |
f46b5a66 | 3507 | |
43d20761 | 3508 | if (!ret) |
0b246afa | 3509 | btrfs_info(fs_info, "disk added %s", vol_args->name); |
43d20761 | 3510 | |
f46b5a66 | 3511 | kfree(vol_args); |
c9e9f97b | 3512 | out: |
a174c0a2 NB |
3513 | if (restore_op) |
3514 | btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED); | |
3515 | else | |
3516 | btrfs_exclop_finish(fs_info); | |
f46b5a66 CH |
3517 | return ret; |
3518 | } | |
3519 | ||
6b526ed7 | 3520 | static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) |
f46b5a66 | 3521 | { |
1a15eb72 | 3522 | BTRFS_DEV_LOOKUP_ARGS(args); |
0b246afa JM |
3523 | struct inode *inode = file_inode(file); |
3524 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
6b526ed7 | 3525 | struct btrfs_ioctl_vol_args_v2 *vol_args; |
3fa421de JB |
3526 | struct block_device *bdev = NULL; |
3527 | fmode_t mode; | |
f46b5a66 | 3528 | int ret; |
67ae34b6 | 3529 | bool cancel = false; |
f46b5a66 | 3530 | |
e441d54d CM |
3531 | if (!capable(CAP_SYS_ADMIN)) |
3532 | return -EPERM; | |
3533 | ||
dae7b665 | 3534 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
d815b3f2 DC |
3535 | if (IS_ERR(vol_args)) |
3536 | return PTR_ERR(vol_args); | |
f46b5a66 | 3537 | |
748449cd | 3538 | if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) { |
fd4e994b OS |
3539 | ret = -EOPNOTSUPP; |
3540 | goto out; | |
3541 | } | |
1a15eb72 | 3542 | |
67ae34b6 | 3543 | vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; |
1a15eb72 JB |
3544 | if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { |
3545 | args.devid = vol_args->devid; | |
3546 | } else if (!strcmp("cancel", vol_args->name)) { | |
67ae34b6 | 3547 | cancel = true; |
1a15eb72 JB |
3548 | } else { |
3549 | ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name); | |
3550 | if (ret) | |
3551 | goto out; | |
3552 | } | |
3553 | ||
3554 | ret = mnt_want_write_file(file); | |
3555 | if (ret) | |
3556 | goto out; | |
f46b5a66 | 3557 | |
67ae34b6 DS |
3558 | ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE, |
3559 | cancel); | |
3560 | if (ret) | |
1a15eb72 | 3561 | goto err_drop; |
183860f6 | 3562 | |
1a15eb72 JB |
3563 | /* Exclusive operation is now claimed */ |
3564 | ret = btrfs_rm_device(fs_info, &args, &bdev, &mode); | |
67ae34b6 | 3565 | |
c3e1f96c | 3566 | btrfs_exclop_finish(fs_info); |
183860f6 | 3567 | |
6b526ed7 | 3568 | if (!ret) { |
735654ea | 3569 | if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) |
0b246afa | 3570 | btrfs_info(fs_info, "device deleted: id %llu", |
6b526ed7 AJ |
3571 | vol_args->devid); |
3572 | else | |
0b246afa | 3573 | btrfs_info(fs_info, "device deleted: %s", |
6b526ed7 AJ |
3574 | vol_args->name); |
3575 | } | |
c47ca32d | 3576 | err_drop: |
4ac20c70 | 3577 | mnt_drop_write_file(file); |
3fa421de JB |
3578 | if (bdev) |
3579 | blkdev_put(bdev, mode); | |
1a15eb72 JB |
3580 | out: |
3581 | btrfs_put_dev_args_from_path(&args); | |
3582 | kfree(vol_args); | |
f46b5a66 CH |
3583 | return ret; |
3584 | } | |
3585 | ||
da24927b | 3586 | static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) |
f46b5a66 | 3587 | { |
1a15eb72 | 3588 | BTRFS_DEV_LOOKUP_ARGS(args); |
0b246afa JM |
3589 | struct inode *inode = file_inode(file); |
3590 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
f46b5a66 | 3591 | struct btrfs_ioctl_vol_args *vol_args; |
3fa421de JB |
3592 | struct block_device *bdev = NULL; |
3593 | fmode_t mode; | |
f46b5a66 | 3594 | int ret; |
37b45995 | 3595 | bool cancel = false; |
f46b5a66 | 3596 | |
e441d54d CM |
3597 | if (!capable(CAP_SYS_ADMIN)) |
3598 | return -EPERM; | |
3599 | ||
58d7bbf8 | 3600 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1a15eb72 JB |
3601 | if (IS_ERR(vol_args)) |
3602 | return PTR_ERR(vol_args); | |
3603 | ||
58d7bbf8 | 3604 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
1a15eb72 JB |
3605 | if (!strcmp("cancel", vol_args->name)) { |
3606 | cancel = true; | |
3607 | } else { | |
3608 | ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name); | |
3609 | if (ret) | |
3610 | goto out; | |
3611 | } | |
3612 | ||
3613 | ret = mnt_want_write_file(file); | |
3614 | if (ret) | |
3615 | goto out; | |
67ae34b6 DS |
3616 | |
3617 | ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE, | |
3618 | cancel); | |
3619 | if (ret == 0) { | |
1a15eb72 | 3620 | ret = btrfs_rm_device(fs_info, &args, &bdev, &mode); |
67ae34b6 DS |
3621 | if (!ret) |
3622 | btrfs_info(fs_info, "disk deleted %s", vol_args->name); | |
3623 | btrfs_exclop_finish(fs_info); | |
3624 | } | |
183860f6 | 3625 | |
4ac20c70 | 3626 | mnt_drop_write_file(file); |
3fa421de JB |
3627 | if (bdev) |
3628 | blkdev_put(bdev, mode); | |
1a15eb72 JB |
3629 | out: |
3630 | btrfs_put_dev_args_from_path(&args); | |
3631 | kfree(vol_args); | |
f46b5a66 CH |
3632 | return ret; |
3633 | } | |
3634 | ||
2ff7e61e JM |
3635 | static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, |
3636 | void __user *arg) | |
475f6387 | 3637 | { |
027ed2f0 | 3638 | struct btrfs_ioctl_fs_info_args *fi_args; |
475f6387 | 3639 | struct btrfs_device *device; |
0b246afa | 3640 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; |
137c5418 | 3641 | u64 flags_in; |
027ed2f0 | 3642 | int ret = 0; |
475f6387 | 3643 | |
137c5418 JT |
3644 | fi_args = memdup_user(arg, sizeof(*fi_args)); |
3645 | if (IS_ERR(fi_args)) | |
3646 | return PTR_ERR(fi_args); | |
3647 | ||
3648 | flags_in = fi_args->flags; | |
3649 | memset(fi_args, 0, sizeof(*fi_args)); | |
027ed2f0 | 3650 | |
d03262c7 | 3651 | rcu_read_lock(); |
027ed2f0 | 3652 | fi_args->num_devices = fs_devices->num_devices; |
475f6387 | 3653 | |
d03262c7 | 3654 | list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { |
027ed2f0 LZ |
3655 | if (device->devid > fi_args->max_id) |
3656 | fi_args->max_id = device->devid; | |
475f6387 | 3657 | } |
d03262c7 | 3658 | rcu_read_unlock(); |
475f6387 | 3659 | |
de37aa51 | 3660 | memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid)); |
bea7eafd OS |
3661 | fi_args->nodesize = fs_info->nodesize; |
3662 | fi_args->sectorsize = fs_info->sectorsize; | |
3663 | fi_args->clone_alignment = fs_info->sectorsize; | |
80a773fb | 3664 | |
137c5418 JT |
3665 | if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) { |
3666 | fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy); | |
3667 | fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy); | |
3668 | fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO; | |
3669 | } | |
3670 | ||
0fb408a5 JT |
3671 | if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) { |
3672 | fi_args->generation = fs_info->generation; | |
3673 | fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION; | |
3674 | } | |
3675 | ||
49bac897 JT |
3676 | if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) { |
3677 | memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid, | |
3678 | sizeof(fi_args->metadata_uuid)); | |
3679 | fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID; | |
3680 | } | |
3681 | ||
027ed2f0 LZ |
3682 | if (copy_to_user(arg, fi_args, sizeof(*fi_args))) |
3683 | ret = -EFAULT; | |
475f6387 | 3684 | |
027ed2f0 LZ |
3685 | kfree(fi_args); |
3686 | return ret; | |
475f6387 JS |
3687 | } |
3688 | ||
2ff7e61e JM |
3689 | static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, |
3690 | void __user *arg) | |
475f6387 | 3691 | { |
562d7b15 | 3692 | BTRFS_DEV_LOOKUP_ARGS(args); |
475f6387 JS |
3693 | struct btrfs_ioctl_dev_info_args *di_args; |
3694 | struct btrfs_device *dev; | |
475f6387 | 3695 | int ret = 0; |
475f6387 | 3696 | |
475f6387 JS |
3697 | di_args = memdup_user(arg, sizeof(*di_args)); |
3698 | if (IS_ERR(di_args)) | |
3699 | return PTR_ERR(di_args); | |
3700 | ||
562d7b15 | 3701 | args.devid = di_args->devid; |
dd5f9615 | 3702 | if (!btrfs_is_empty_uuid(di_args->uuid)) |
562d7b15 | 3703 | args.uuid = di_args->uuid; |
475f6387 | 3704 | |
c5593ca3 | 3705 | rcu_read_lock(); |
562d7b15 | 3706 | dev = btrfs_find_device(fs_info->fs_devices, &args); |
475f6387 JS |
3707 | if (!dev) { |
3708 | ret = -ENODEV; | |
3709 | goto out; | |
3710 | } | |
3711 | ||
3712 | di_args->devid = dev->devid; | |
7cc8e58d MX |
3713 | di_args->bytes_used = btrfs_device_get_bytes_used(dev); |
3714 | di_args->total_bytes = btrfs_device_get_total_bytes(dev); | |
475f6387 | 3715 | memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); |
a27202fb | 3716 | if (dev->name) { |
672d5990 MT |
3717 | strncpy(di_args->path, rcu_str_deref(dev->name), |
3718 | sizeof(di_args->path) - 1); | |
a27202fb JM |
3719 | di_args->path[sizeof(di_args->path) - 1] = 0; |
3720 | } else { | |
99ba55ad | 3721 | di_args->path[0] = '\0'; |
a27202fb | 3722 | } |
475f6387 JS |
3723 | |
3724 | out: | |
c5593ca3 | 3725 | rcu_read_unlock(); |
475f6387 JS |
3726 | if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) |
3727 | ret = -EFAULT; | |
3728 | ||
3729 | kfree(di_args); | |
3730 | return ret; | |
3731 | } | |
3732 | ||
6ef5ed0d JB |
3733 | static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) |
3734 | { | |
496ad9aa | 3735 | struct inode *inode = file_inode(file); |
0b246afa | 3736 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
6ef5ed0d JB |
3737 | struct btrfs_root *root = BTRFS_I(inode)->root; |
3738 | struct btrfs_root *new_root; | |
3739 | struct btrfs_dir_item *di; | |
3740 | struct btrfs_trans_handle *trans; | |
2a2b5d62 | 3741 | struct btrfs_path *path = NULL; |
6ef5ed0d | 3742 | struct btrfs_disk_key disk_key; |
6ef5ed0d JB |
3743 | u64 objectid = 0; |
3744 | u64 dir_id; | |
3c04ce01 | 3745 | int ret; |
6ef5ed0d JB |
3746 | |
3747 | if (!capable(CAP_SYS_ADMIN)) | |
3748 | return -EPERM; | |
3749 | ||
3c04ce01 MX |
3750 | ret = mnt_want_write_file(file); |
3751 | if (ret) | |
3752 | return ret; | |
3753 | ||
3754 | if (copy_from_user(&objectid, argp, sizeof(objectid))) { | |
3755 | ret = -EFAULT; | |
3756 | goto out; | |
3757 | } | |
6ef5ed0d JB |
3758 | |
3759 | if (!objectid) | |
1cecf579 | 3760 | objectid = BTRFS_FS_TREE_OBJECTID; |
6ef5ed0d | 3761 | |
56e9357a | 3762 | new_root = btrfs_get_fs_root(fs_info, objectid, true); |
3c04ce01 MX |
3763 | if (IS_ERR(new_root)) { |
3764 | ret = PTR_ERR(new_root); | |
3765 | goto out; | |
3766 | } | |
2a2b5d62 JB |
3767 | if (!is_fstree(new_root->root_key.objectid)) { |
3768 | ret = -ENOENT; | |
3769 | goto out_free; | |
3770 | } | |
6ef5ed0d | 3771 | |
6ef5ed0d | 3772 | path = btrfs_alloc_path(); |
3c04ce01 MX |
3773 | if (!path) { |
3774 | ret = -ENOMEM; | |
2a2b5d62 | 3775 | goto out_free; |
3c04ce01 | 3776 | } |
6ef5ed0d JB |
3777 | |
3778 | trans = btrfs_start_transaction(root, 1); | |
98d5dc13 | 3779 | if (IS_ERR(trans)) { |
3c04ce01 | 3780 | ret = PTR_ERR(trans); |
2a2b5d62 | 3781 | goto out_free; |
6ef5ed0d JB |
3782 | } |
3783 | ||
0b246afa JM |
3784 | dir_id = btrfs_super_root_dir(fs_info->super_copy); |
3785 | di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, | |
6ef5ed0d | 3786 | dir_id, "default", 7, 1); |
cf1e99a4 | 3787 | if (IS_ERR_OR_NULL(di)) { |
2a2b5d62 | 3788 | btrfs_release_path(path); |
3a45bb20 | 3789 | btrfs_end_transaction(trans); |
0b246afa | 3790 | btrfs_err(fs_info, |
5d163e0e | 3791 | "Umm, you don't have the default diritem, this isn't going to work"); |
3c04ce01 | 3792 | ret = -ENOENT; |
2a2b5d62 | 3793 | goto out_free; |
6ef5ed0d JB |
3794 | } |
3795 | ||
3796 | btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); | |
3797 | btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); | |
3798 | btrfs_mark_buffer_dirty(path->nodes[0]); | |
2a2b5d62 | 3799 | btrfs_release_path(path); |
6ef5ed0d | 3800 | |
0b246afa | 3801 | btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); |
3a45bb20 | 3802 | btrfs_end_transaction(trans); |
2a2b5d62 | 3803 | out_free: |
00246528 | 3804 | btrfs_put_root(new_root); |
2a2b5d62 | 3805 | btrfs_free_path(path); |
3c04ce01 MX |
3806 | out: |
3807 | mnt_drop_write_file(file); | |
3808 | return ret; | |
6ef5ed0d JB |
3809 | } |
3810 | ||
c065f5b1 SY |
3811 | static void get_block_group_info(struct list_head *groups_list, |
3812 | struct btrfs_ioctl_space_info *space) | |
bf5fc093 | 3813 | { |
32da5386 | 3814 | struct btrfs_block_group *block_group; |
bf5fc093 JB |
3815 | |
3816 | space->total_bytes = 0; | |
3817 | space->used_bytes = 0; | |
3818 | space->flags = 0; | |
3819 | list_for_each_entry(block_group, groups_list, list) { | |
3820 | space->flags = block_group->flags; | |
b3470b5d | 3821 | space->total_bytes += block_group->length; |
bf38be65 | 3822 | space->used_bytes += block_group->used; |
bf5fc093 JB |
3823 | } |
3824 | } | |
3825 | ||
2ff7e61e JM |
3826 | static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, |
3827 | void __user *arg) | |
1406e432 JB |
3828 | { |
3829 | struct btrfs_ioctl_space_args space_args; | |
3830 | struct btrfs_ioctl_space_info space; | |
3831 | struct btrfs_ioctl_space_info *dest; | |
7fde62bf | 3832 | struct btrfs_ioctl_space_info *dest_orig; |
13f2696f | 3833 | struct btrfs_ioctl_space_info __user *user_dest; |
1406e432 | 3834 | struct btrfs_space_info *info; |
315d8e98 CIK |
3835 | static const u64 types[] = { |
3836 | BTRFS_BLOCK_GROUP_DATA, | |
3837 | BTRFS_BLOCK_GROUP_SYSTEM, | |
3838 | BTRFS_BLOCK_GROUP_METADATA, | |
3839 | BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA | |
3840 | }; | |
bf5fc093 | 3841 | int num_types = 4; |
7fde62bf | 3842 | int alloc_size; |
1406e432 | 3843 | int ret = 0; |
51788b1b | 3844 | u64 slot_count = 0; |
bf5fc093 | 3845 | int i, c; |
1406e432 JB |
3846 | |
3847 | if (copy_from_user(&space_args, | |
3848 | (struct btrfs_ioctl_space_args __user *)arg, | |
3849 | sizeof(space_args))) | |
3850 | return -EFAULT; | |
3851 | ||
bf5fc093 JB |
3852 | for (i = 0; i < num_types; i++) { |
3853 | struct btrfs_space_info *tmp; | |
3854 | ||
3855 | info = NULL; | |
72804905 | 3856 | list_for_each_entry(tmp, &fs_info->space_info, list) { |
bf5fc093 JB |
3857 | if (tmp->flags == types[i]) { |
3858 | info = tmp; | |
3859 | break; | |
3860 | } | |
3861 | } | |
bf5fc093 JB |
3862 | |
3863 | if (!info) | |
3864 | continue; | |
3865 | ||
3866 | down_read(&info->groups_sem); | |
3867 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
3868 | if (!list_empty(&info->block_groups[c])) | |
3869 | slot_count++; | |
3870 | } | |
3871 | up_read(&info->groups_sem); | |
3872 | } | |
7fde62bf | 3873 | |
36523e95 DS |
3874 | /* |
3875 | * Global block reserve, exported as a space_info | |
3876 | */ | |
3877 | slot_count++; | |
3878 | ||
7fde62bf CM |
3879 | /* space_slots == 0 means they are asking for a count */ |
3880 | if (space_args.space_slots == 0) { | |
3881 | space_args.total_spaces = slot_count; | |
3882 | goto out; | |
3883 | } | |
bf5fc093 | 3884 | |
51788b1b | 3885 | slot_count = min_t(u64, space_args.space_slots, slot_count); |
bf5fc093 | 3886 | |
7fde62bf | 3887 | alloc_size = sizeof(*dest) * slot_count; |
bf5fc093 | 3888 | |
7fde62bf CM |
3889 | /* we generally have at most 6 or so space infos, one for each raid |
3890 | * level. So, a whole page should be more than enough for everyone | |
3891 | */ | |
09cbfeaf | 3892 | if (alloc_size > PAGE_SIZE) |
7fde62bf CM |
3893 | return -ENOMEM; |
3894 | ||
1406e432 | 3895 | space_args.total_spaces = 0; |
8d2db785 | 3896 | dest = kmalloc(alloc_size, GFP_KERNEL); |
7fde62bf CM |
3897 | if (!dest) |
3898 | return -ENOMEM; | |
3899 | dest_orig = dest; | |
1406e432 | 3900 | |
7fde62bf | 3901 | /* now we have a buffer to copy into */ |
bf5fc093 JB |
3902 | for (i = 0; i < num_types; i++) { |
3903 | struct btrfs_space_info *tmp; | |
3904 | ||
51788b1b DR |
3905 | if (!slot_count) |
3906 | break; | |
3907 | ||
bf5fc093 | 3908 | info = NULL; |
72804905 | 3909 | list_for_each_entry(tmp, &fs_info->space_info, list) { |
bf5fc093 JB |
3910 | if (tmp->flags == types[i]) { |
3911 | info = tmp; | |
3912 | break; | |
3913 | } | |
3914 | } | |
7fde62bf | 3915 | |
bf5fc093 JB |
3916 | if (!info) |
3917 | continue; | |
3918 | down_read(&info->groups_sem); | |
3919 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
3920 | if (!list_empty(&info->block_groups[c])) { | |
c065f5b1 SY |
3921 | get_block_group_info(&info->block_groups[c], |
3922 | &space); | |
bf5fc093 JB |
3923 | memcpy(dest, &space, sizeof(space)); |
3924 | dest++; | |
3925 | space_args.total_spaces++; | |
51788b1b | 3926 | slot_count--; |
bf5fc093 | 3927 | } |
51788b1b DR |
3928 | if (!slot_count) |
3929 | break; | |
bf5fc093 JB |
3930 | } |
3931 | up_read(&info->groups_sem); | |
1406e432 | 3932 | } |
1406e432 | 3933 | |
36523e95 DS |
3934 | /* |
3935 | * Add global block reserve | |
3936 | */ | |
3937 | if (slot_count) { | |
0b246afa | 3938 | struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; |
36523e95 DS |
3939 | |
3940 | spin_lock(&block_rsv->lock); | |
3941 | space.total_bytes = block_rsv->size; | |
3942 | space.used_bytes = block_rsv->size - block_rsv->reserved; | |
3943 | spin_unlock(&block_rsv->lock); | |
3944 | space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; | |
3945 | memcpy(dest, &space, sizeof(space)); | |
3946 | space_args.total_spaces++; | |
3947 | } | |
3948 | ||
2eec6c81 | 3949 | user_dest = (struct btrfs_ioctl_space_info __user *) |
7fde62bf CM |
3950 | (arg + sizeof(struct btrfs_ioctl_space_args)); |
3951 | ||
3952 | if (copy_to_user(user_dest, dest_orig, alloc_size)) | |
3953 | ret = -EFAULT; | |
3954 | ||
3955 | kfree(dest_orig); | |
3956 | out: | |
3957 | if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) | |
1406e432 JB |
3958 | ret = -EFAULT; |
3959 | ||
3960 | return ret; | |
3961 | } | |
3962 | ||
9a8c28be MX |
3963 | static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, |
3964 | void __user *argp) | |
46204592 | 3965 | { |
46204592 SW |
3966 | struct btrfs_trans_handle *trans; |
3967 | u64 transid; | |
3968 | ||
d4edf39b | 3969 | trans = btrfs_attach_transaction_barrier(root); |
ff7c1d33 MX |
3970 | if (IS_ERR(trans)) { |
3971 | if (PTR_ERR(trans) != -ENOENT) | |
3972 | return PTR_ERR(trans); | |
3973 | ||
3974 | /* No running transaction, don't bother */ | |
3975 | transid = root->fs_info->last_trans_committed; | |
3976 | goto out; | |
3977 | } | |
46204592 | 3978 | transid = trans->transid; |
fdfbf020 | 3979 | btrfs_commit_transaction_async(trans); |
ff7c1d33 | 3980 | out: |
46204592 SW |
3981 | if (argp) |
3982 | if (copy_to_user(argp, &transid, sizeof(transid))) | |
3983 | return -EFAULT; | |
3984 | return 0; | |
3985 | } | |
3986 | ||
2ff7e61e | 3987 | static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, |
9a8c28be | 3988 | void __user *argp) |
46204592 | 3989 | { |
46204592 SW |
3990 | u64 transid; |
3991 | ||
3992 | if (argp) { | |
3993 | if (copy_from_user(&transid, argp, sizeof(transid))) | |
3994 | return -EFAULT; | |
3995 | } else { | |
3996 | transid = 0; /* current trans */ | |
3997 | } | |
2ff7e61e | 3998 | return btrfs_wait_for_commit(fs_info, transid); |
46204592 SW |
3999 | } |
4000 | ||
b8e95489 | 4001 | static long btrfs_ioctl_scrub(struct file *file, void __user *arg) |
475f6387 | 4002 | { |
0b246afa | 4003 | struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); |
475f6387 | 4004 | struct btrfs_ioctl_scrub_args *sa; |
b8e95489 | 4005 | int ret; |
475f6387 JS |
4006 | |
4007 | if (!capable(CAP_SYS_ADMIN)) | |
4008 | return -EPERM; | |
4009 | ||
da32c6d5 JB |
4010 | if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { |
4011 | btrfs_err(fs_info, "scrub is not supported on extent tree v2 yet"); | |
4012 | return -EINVAL; | |
4013 | } | |
4014 | ||
475f6387 JS |
4015 | sa = memdup_user(arg, sizeof(*sa)); |
4016 | if (IS_ERR(sa)) | |
4017 | return PTR_ERR(sa); | |
4018 | ||
b8e95489 MX |
4019 | if (!(sa->flags & BTRFS_SCRUB_READONLY)) { |
4020 | ret = mnt_want_write_file(file); | |
4021 | if (ret) | |
4022 | goto out; | |
4023 | } | |
4024 | ||
0b246afa | 4025 | ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, |
63a212ab SB |
4026 | &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, |
4027 | 0); | |
475f6387 | 4028 | |
5afe6ce7 FM |
4029 | /* |
4030 | * Copy scrub args to user space even if btrfs_scrub_dev() returned an | |
4031 | * error. This is important as it allows user space to know how much | |
4032 | * progress scrub has done. For example, if scrub is canceled we get | |
4033 | * -ECANCELED from btrfs_scrub_dev() and return that error back to user | |
4034 | * space. Later user space can inspect the progress from the structure | |
4035 | * btrfs_ioctl_scrub_args and resume scrub from where it left off | |
4036 | * previously (btrfs-progs does this). | |
4037 | * If we fail to copy the btrfs_ioctl_scrub_args structure to user space | |
4038 | * then return -EFAULT to signal the structure was not copied or it may | |
4039 | * be corrupt and unreliable due to a partial copy. | |
4040 | */ | |
4041 | if (copy_to_user(arg, sa, sizeof(*sa))) | |
475f6387 JS |
4042 | ret = -EFAULT; |
4043 | ||
b8e95489 MX |
4044 | if (!(sa->flags & BTRFS_SCRUB_READONLY)) |
4045 | mnt_drop_write_file(file); | |
4046 | out: | |
475f6387 JS |
4047 | kfree(sa); |
4048 | return ret; | |
4049 | } | |
4050 | ||
2ff7e61e | 4051 | static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) |
475f6387 JS |
4052 | { |
4053 | if (!capable(CAP_SYS_ADMIN)) | |
4054 | return -EPERM; | |
4055 | ||
2ff7e61e | 4056 | return btrfs_scrub_cancel(fs_info); |
475f6387 JS |
4057 | } |
4058 | ||
2ff7e61e | 4059 | static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, |
475f6387 JS |
4060 | void __user *arg) |
4061 | { | |
4062 | struct btrfs_ioctl_scrub_args *sa; | |
4063 | int ret; | |
4064 | ||
4065 | if (!capable(CAP_SYS_ADMIN)) | |
4066 | return -EPERM; | |
4067 | ||
4068 | sa = memdup_user(arg, sizeof(*sa)); | |
4069 | if (IS_ERR(sa)) | |
4070 | return PTR_ERR(sa); | |
4071 | ||
2ff7e61e | 4072 | ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); |
475f6387 | 4073 | |
4fa99b00 | 4074 | if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) |
475f6387 JS |
4075 | ret = -EFAULT; |
4076 | ||
4077 | kfree(sa); | |
4078 | return ret; | |
4079 | } | |
4080 | ||
2ff7e61e | 4081 | static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, |
b27f7c0c | 4082 | void __user *arg) |
c11d2c23 SB |
4083 | { |
4084 | struct btrfs_ioctl_get_dev_stats *sa; | |
4085 | int ret; | |
4086 | ||
c11d2c23 SB |
4087 | sa = memdup_user(arg, sizeof(*sa)); |
4088 | if (IS_ERR(sa)) | |
4089 | return PTR_ERR(sa); | |
4090 | ||
b27f7c0c DS |
4091 | if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { |
4092 | kfree(sa); | |
4093 | return -EPERM; | |
4094 | } | |
4095 | ||
2ff7e61e | 4096 | ret = btrfs_get_dev_stats(fs_info, sa); |
c11d2c23 | 4097 | |
eee99577 | 4098 | if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) |
c11d2c23 SB |
4099 | ret = -EFAULT; |
4100 | ||
4101 | kfree(sa); | |
4102 | return ret; | |
4103 | } | |
4104 | ||
2ff7e61e JM |
4105 | static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, |
4106 | void __user *arg) | |
3f6bcfbd SB |
4107 | { |
4108 | struct btrfs_ioctl_dev_replace_args *p; | |
4109 | int ret; | |
4110 | ||
4111 | if (!capable(CAP_SYS_ADMIN)) | |
4112 | return -EPERM; | |
4113 | ||
914a519b JB |
4114 | if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { |
4115 | btrfs_err(fs_info, "device replace not supported on extent tree v2 yet"); | |
4116 | return -EINVAL; | |
4117 | } | |
4118 | ||
3f6bcfbd SB |
4119 | p = memdup_user(arg, sizeof(*p)); |
4120 | if (IS_ERR(p)) | |
4121 | return PTR_ERR(p); | |
4122 | ||
4123 | switch (p->cmd) { | |
4124 | case BTRFS_IOCTL_DEV_REPLACE_CMD_START: | |
bc98a42c | 4125 | if (sb_rdonly(fs_info->sb)) { |
adfa97cb ID |
4126 | ret = -EROFS; |
4127 | goto out; | |
4128 | } | |
c3e1f96c | 4129 | if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) { |
e57138b3 | 4130 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
3f6bcfbd | 4131 | } else { |
2ff7e61e | 4132 | ret = btrfs_dev_replace_by_ioctl(fs_info, p); |
c3e1f96c | 4133 | btrfs_exclop_finish(fs_info); |
3f6bcfbd SB |
4134 | } |
4135 | break; | |
4136 | case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: | |
0b246afa | 4137 | btrfs_dev_replace_status(fs_info, p); |
3f6bcfbd SB |
4138 | ret = 0; |
4139 | break; | |
4140 | case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: | |
17d202b9 | 4141 | p->result = btrfs_dev_replace_cancel(fs_info); |
97282031 | 4142 | ret = 0; |
3f6bcfbd SB |
4143 | break; |
4144 | default: | |
4145 | ret = -EINVAL; | |
4146 | break; | |
4147 | } | |
4148 | ||
d3a53286 | 4149 | if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p))) |
3f6bcfbd | 4150 | ret = -EFAULT; |
adfa97cb | 4151 | out: |
3f6bcfbd SB |
4152 | kfree(p); |
4153 | return ret; | |
4154 | } | |
4155 | ||
d7728c96 JS |
4156 | static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) |
4157 | { | |
4158 | int ret = 0; | |
4159 | int i; | |
740c3d22 | 4160 | u64 rel_ptr; |
d7728c96 | 4161 | int size; |
806468f8 | 4162 | struct btrfs_ioctl_ino_path_args *ipa = NULL; |
d7728c96 JS |
4163 | struct inode_fs_paths *ipath = NULL; |
4164 | struct btrfs_path *path; | |
4165 | ||
82b22ac8 | 4166 | if (!capable(CAP_DAC_READ_SEARCH)) |
d7728c96 JS |
4167 | return -EPERM; |
4168 | ||
4169 | path = btrfs_alloc_path(); | |
4170 | if (!path) { | |
4171 | ret = -ENOMEM; | |
4172 | goto out; | |
4173 | } | |
4174 | ||
4175 | ipa = memdup_user(arg, sizeof(*ipa)); | |
4176 | if (IS_ERR(ipa)) { | |
4177 | ret = PTR_ERR(ipa); | |
4178 | ipa = NULL; | |
4179 | goto out; | |
4180 | } | |
4181 | ||
4182 | size = min_t(u32, ipa->size, 4096); | |
4183 | ipath = init_ipath(size, root, path); | |
4184 | if (IS_ERR(ipath)) { | |
4185 | ret = PTR_ERR(ipath); | |
4186 | ipath = NULL; | |
4187 | goto out; | |
4188 | } | |
4189 | ||
4190 | ret = paths_from_inode(ipa->inum, ipath); | |
4191 | if (ret < 0) | |
4192 | goto out; | |
4193 | ||
4194 | for (i = 0; i < ipath->fspath->elem_cnt; ++i) { | |
745c4d8e JM |
4195 | rel_ptr = ipath->fspath->val[i] - |
4196 | (u64)(unsigned long)ipath->fspath->val; | |
740c3d22 | 4197 | ipath->fspath->val[i] = rel_ptr; |
d7728c96 JS |
4198 | } |
4199 | ||
718dc5fa OS |
4200 | ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, |
4201 | ipath->fspath, size); | |
d7728c96 JS |
4202 | if (ret) { |
4203 | ret = -EFAULT; | |
4204 | goto out; | |
4205 | } | |
4206 | ||
4207 | out: | |
4208 | btrfs_free_path(path); | |
4209 | free_ipath(ipath); | |
4210 | kfree(ipa); | |
4211 | ||
4212 | return ret; | |
4213 | } | |
4214 | ||
4215 | static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) | |
4216 | { | |
4217 | struct btrfs_data_container *inodes = ctx; | |
4218 | const size_t c = 3 * sizeof(u64); | |
4219 | ||
4220 | if (inodes->bytes_left >= c) { | |
4221 | inodes->bytes_left -= c; | |
4222 | inodes->val[inodes->elem_cnt] = inum; | |
4223 | inodes->val[inodes->elem_cnt + 1] = offset; | |
4224 | inodes->val[inodes->elem_cnt + 2] = root; | |
4225 | inodes->elem_cnt += 3; | |
4226 | } else { | |
4227 | inodes->bytes_missing += c - inodes->bytes_left; | |
4228 | inodes->bytes_left = 0; | |
4229 | inodes->elem_missed += 3; | |
4230 | } | |
4231 | ||
4232 | return 0; | |
4233 | } | |
4234 | ||
2ff7e61e | 4235 | static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, |
d24a67b2 | 4236 | void __user *arg, int version) |
d7728c96 JS |
4237 | { |
4238 | int ret = 0; | |
4239 | int size; | |
d7728c96 JS |
4240 | struct btrfs_ioctl_logical_ino_args *loi; |
4241 | struct btrfs_data_container *inodes = NULL; | |
4242 | struct btrfs_path *path = NULL; | |
d24a67b2 | 4243 | bool ignore_offset; |
d7728c96 JS |
4244 | |
4245 | if (!capable(CAP_SYS_ADMIN)) | |
4246 | return -EPERM; | |
4247 | ||
4248 | loi = memdup_user(arg, sizeof(*loi)); | |
7b9ea627 SV |
4249 | if (IS_ERR(loi)) |
4250 | return PTR_ERR(loi); | |
d7728c96 | 4251 | |
d24a67b2 ZB |
4252 | if (version == 1) { |
4253 | ignore_offset = false; | |
b115e3bc | 4254 | size = min_t(u32, loi->size, SZ_64K); |
d24a67b2 ZB |
4255 | } else { |
4256 | /* All reserved bits must be 0 for now */ | |
4257 | if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) { | |
4258 | ret = -EINVAL; | |
4259 | goto out_loi; | |
4260 | } | |
4261 | /* Only accept flags we have defined so far */ | |
4262 | if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) { | |
4263 | ret = -EINVAL; | |
4264 | goto out_loi; | |
4265 | } | |
4266 | ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; | |
b115e3bc | 4267 | size = min_t(u32, loi->size, SZ_16M); |
d24a67b2 ZB |
4268 | } |
4269 | ||
d7728c96 JS |
4270 | path = btrfs_alloc_path(); |
4271 | if (!path) { | |
4272 | ret = -ENOMEM; | |
4273 | goto out; | |
4274 | } | |
4275 | ||
d7728c96 JS |
4276 | inodes = init_data_container(size); |
4277 | if (IS_ERR(inodes)) { | |
4278 | ret = PTR_ERR(inodes); | |
4279 | inodes = NULL; | |
4280 | goto out; | |
4281 | } | |
4282 | ||
2ff7e61e | 4283 | ret = iterate_inodes_from_logical(loi->logical, fs_info, path, |
d24a67b2 | 4284 | build_ino_list, inodes, ignore_offset); |
df031f07 | 4285 | if (ret == -EINVAL) |
d7728c96 JS |
4286 | ret = -ENOENT; |
4287 | if (ret < 0) | |
4288 | goto out; | |
4289 | ||
718dc5fa OS |
4290 | ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, |
4291 | size); | |
d7728c96 JS |
4292 | if (ret) |
4293 | ret = -EFAULT; | |
4294 | ||
4295 | out: | |
4296 | btrfs_free_path(path); | |
f54de068 | 4297 | kvfree(inodes); |
d24a67b2 | 4298 | out_loi: |
d7728c96 JS |
4299 | kfree(loi); |
4300 | ||
4301 | return ret; | |
4302 | } | |
4303 | ||
008ef096 | 4304 | void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, |
c9e9f97b ID |
4305 | struct btrfs_ioctl_balance_args *bargs) |
4306 | { | |
4307 | struct btrfs_balance_control *bctl = fs_info->balance_ctl; | |
4308 | ||
4309 | bargs->flags = bctl->flags; | |
4310 | ||
3009a62f | 4311 | if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) |
837d5b6e ID |
4312 | bargs->state |= BTRFS_BALANCE_STATE_RUNNING; |
4313 | if (atomic_read(&fs_info->balance_pause_req)) | |
4314 | bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; | |
a7e99c69 ID |
4315 | if (atomic_read(&fs_info->balance_cancel_req)) |
4316 | bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; | |
837d5b6e | 4317 | |
c9e9f97b ID |
4318 | memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); |
4319 | memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); | |
4320 | memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); | |
19a39dce | 4321 | |
008ef096 DS |
4322 | spin_lock(&fs_info->balance_lock); |
4323 | memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); | |
4324 | spin_unlock(&fs_info->balance_lock); | |
c9e9f97b ID |
4325 | } |
4326 | ||
9ba1f6e4 | 4327 | static long btrfs_ioctl_balance(struct file *file, void __user *arg) |
c9e9f97b | 4328 | { |
496ad9aa | 4329 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
c9e9f97b ID |
4330 | struct btrfs_fs_info *fs_info = root->fs_info; |
4331 | struct btrfs_ioctl_balance_args *bargs; | |
4332 | struct btrfs_balance_control *bctl; | |
ed0fb78f | 4333 | bool need_unlock; /* for mut. excl. ops lock */ |
c9e9f97b ID |
4334 | int ret; |
4335 | ||
6c405b24 NB |
4336 | if (!arg) |
4337 | btrfs_warn(fs_info, | |
4338 | "IOC_BALANCE ioctl (v1) is deprecated and will be removed in kernel 5.18"); | |
4339 | ||
c9e9f97b ID |
4340 | if (!capable(CAP_SYS_ADMIN)) |
4341 | return -EPERM; | |
4342 | ||
e54bfa31 | 4343 | ret = mnt_want_write_file(file); |
9ba1f6e4 LB |
4344 | if (ret) |
4345 | return ret; | |
4346 | ||
ed0fb78f | 4347 | again: |
c3e1f96c | 4348 | if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) { |
ed0fb78f ID |
4349 | mutex_lock(&fs_info->balance_mutex); |
4350 | need_unlock = true; | |
4351 | goto locked; | |
4352 | } | |
4353 | ||
4354 | /* | |
01327610 | 4355 | * mut. excl. ops lock is locked. Three possibilities: |
ed0fb78f ID |
4356 | * (1) some other op is running |
4357 | * (2) balance is running | |
4358 | * (3) balance is paused -- special case (think resume) | |
4359 | */ | |
c9e9f97b | 4360 | mutex_lock(&fs_info->balance_mutex); |
ed0fb78f ID |
4361 | if (fs_info->balance_ctl) { |
4362 | /* this is either (2) or (3) */ | |
3009a62f | 4363 | if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { |
ed0fb78f | 4364 | mutex_unlock(&fs_info->balance_mutex); |
dccdb07b DS |
4365 | /* |
4366 | * Lock released to allow other waiters to continue, | |
4367 | * we'll reexamine the status again. | |
4368 | */ | |
ed0fb78f ID |
4369 | mutex_lock(&fs_info->balance_mutex); |
4370 | ||
4371 | if (fs_info->balance_ctl && | |
3009a62f | 4372 | !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { |
ed0fb78f ID |
4373 | /* this is (3) */ |
4374 | need_unlock = false; | |
4375 | goto locked; | |
4376 | } | |
4377 | ||
4378 | mutex_unlock(&fs_info->balance_mutex); | |
ed0fb78f ID |
4379 | goto again; |
4380 | } else { | |
4381 | /* this is (2) */ | |
4382 | mutex_unlock(&fs_info->balance_mutex); | |
4383 | ret = -EINPROGRESS; | |
4384 | goto out; | |
4385 | } | |
4386 | } else { | |
4387 | /* this is (1) */ | |
4388 | mutex_unlock(&fs_info->balance_mutex); | |
e57138b3 | 4389 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
ed0fb78f ID |
4390 | goto out; |
4391 | } | |
4392 | ||
4393 | locked: | |
c9e9f97b ID |
4394 | |
4395 | if (arg) { | |
4396 | bargs = memdup_user(arg, sizeof(*bargs)); | |
4397 | if (IS_ERR(bargs)) { | |
4398 | ret = PTR_ERR(bargs); | |
ed0fb78f | 4399 | goto out_unlock; |
c9e9f97b | 4400 | } |
de322263 ID |
4401 | |
4402 | if (bargs->flags & BTRFS_BALANCE_RESUME) { | |
4403 | if (!fs_info->balance_ctl) { | |
4404 | ret = -ENOTCONN; | |
4405 | goto out_bargs; | |
4406 | } | |
4407 | ||
4408 | bctl = fs_info->balance_ctl; | |
4409 | spin_lock(&fs_info->balance_lock); | |
4410 | bctl->flags |= BTRFS_BALANCE_RESUME; | |
4411 | spin_unlock(&fs_info->balance_lock); | |
efc0e69c | 4412 | btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE); |
de322263 ID |
4413 | |
4414 | goto do_balance; | |
4415 | } | |
c9e9f97b ID |
4416 | } else { |
4417 | bargs = NULL; | |
4418 | } | |
4419 | ||
ed0fb78f | 4420 | if (fs_info->balance_ctl) { |
837d5b6e ID |
4421 | ret = -EINPROGRESS; |
4422 | goto out_bargs; | |
4423 | } | |
4424 | ||
8d2db785 | 4425 | bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); |
c9e9f97b ID |
4426 | if (!bctl) { |
4427 | ret = -ENOMEM; | |
4428 | goto out_bargs; | |
4429 | } | |
4430 | ||
c9e9f97b ID |
4431 | if (arg) { |
4432 | memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); | |
4433 | memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); | |
4434 | memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); | |
4435 | ||
4436 | bctl->flags = bargs->flags; | |
f43ffb60 ID |
4437 | } else { |
4438 | /* balance everything - no filters */ | |
4439 | bctl->flags |= BTRFS_BALANCE_TYPE_MASK; | |
c9e9f97b ID |
4440 | } |
4441 | ||
8eb93459 DS |
4442 | if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { |
4443 | ret = -EINVAL; | |
0f89abf5 | 4444 | goto out_bctl; |
8eb93459 DS |
4445 | } |
4446 | ||
de322263 | 4447 | do_balance: |
c9e9f97b | 4448 | /* |
c3e1f96c GR |
4449 | * Ownership of bctl and exclusive operation goes to btrfs_balance. |
4450 | * bctl is freed in reset_balance_state, or, if restriper was paused | |
4451 | * all the way until unmount, in free_fs_info. The flag should be | |
4452 | * cleared after reset_balance_state. | |
c9e9f97b | 4453 | */ |
ed0fb78f ID |
4454 | need_unlock = false; |
4455 | ||
6fcf6e2b | 4456 | ret = btrfs_balance(fs_info, bctl, bargs); |
0f89abf5 | 4457 | bctl = NULL; |
ed0fb78f | 4458 | |
d00c2d9c | 4459 | if ((ret == 0 || ret == -ECANCELED) && arg) { |
c9e9f97b ID |
4460 | if (copy_to_user(arg, bargs, sizeof(*bargs))) |
4461 | ret = -EFAULT; | |
4462 | } | |
4463 | ||
0f89abf5 CE |
4464 | out_bctl: |
4465 | kfree(bctl); | |
c9e9f97b ID |
4466 | out_bargs: |
4467 | kfree(bargs); | |
ed0fb78f | 4468 | out_unlock: |
c9e9f97b | 4469 | mutex_unlock(&fs_info->balance_mutex); |
ed0fb78f | 4470 | if (need_unlock) |
c3e1f96c | 4471 | btrfs_exclop_finish(fs_info); |
ed0fb78f | 4472 | out: |
e54bfa31 | 4473 | mnt_drop_write_file(file); |
c9e9f97b ID |
4474 | return ret; |
4475 | } | |
4476 | ||
2ff7e61e | 4477 | static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) |
837d5b6e ID |
4478 | { |
4479 | if (!capable(CAP_SYS_ADMIN)) | |
4480 | return -EPERM; | |
4481 | ||
4482 | switch (cmd) { | |
4483 | case BTRFS_BALANCE_CTL_PAUSE: | |
0b246afa | 4484 | return btrfs_pause_balance(fs_info); |
a7e99c69 | 4485 | case BTRFS_BALANCE_CTL_CANCEL: |
0b246afa | 4486 | return btrfs_cancel_balance(fs_info); |
837d5b6e ID |
4487 | } |
4488 | ||
4489 | return -EINVAL; | |
4490 | } | |
4491 | ||
2ff7e61e | 4492 | static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, |
19a39dce ID |
4493 | void __user *arg) |
4494 | { | |
19a39dce ID |
4495 | struct btrfs_ioctl_balance_args *bargs; |
4496 | int ret = 0; | |
4497 | ||
4498 | if (!capable(CAP_SYS_ADMIN)) | |
4499 | return -EPERM; | |
4500 | ||
4501 | mutex_lock(&fs_info->balance_mutex); | |
4502 | if (!fs_info->balance_ctl) { | |
4503 | ret = -ENOTCONN; | |
4504 | goto out; | |
4505 | } | |
4506 | ||
8d2db785 | 4507 | bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); |
19a39dce ID |
4508 | if (!bargs) { |
4509 | ret = -ENOMEM; | |
4510 | goto out; | |
4511 | } | |
4512 | ||
008ef096 | 4513 | btrfs_update_ioctl_balance_args(fs_info, bargs); |
19a39dce ID |
4514 | |
4515 | if (copy_to_user(arg, bargs, sizeof(*bargs))) | |
4516 | ret = -EFAULT; | |
4517 | ||
4518 | kfree(bargs); | |
4519 | out: | |
4520 | mutex_unlock(&fs_info->balance_mutex); | |
4521 | return ret; | |
4522 | } | |
4523 | ||
905b0dda | 4524 | static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) |
5d13a37b | 4525 | { |
0b246afa JM |
4526 | struct inode *inode = file_inode(file); |
4527 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
5d13a37b | 4528 | struct btrfs_ioctl_quota_ctl_args *sa; |
5d13a37b | 4529 | int ret; |
5d13a37b AJ |
4530 | |
4531 | if (!capable(CAP_SYS_ADMIN)) | |
4532 | return -EPERM; | |
4533 | ||
905b0dda MX |
4534 | ret = mnt_want_write_file(file); |
4535 | if (ret) | |
4536 | return ret; | |
5d13a37b AJ |
4537 | |
4538 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4539 | if (IS_ERR(sa)) { |
4540 | ret = PTR_ERR(sa); | |
4541 | goto drop_write; | |
4542 | } | |
5d13a37b | 4543 | |
0b246afa | 4544 | down_write(&fs_info->subvol_sem); |
5d13a37b AJ |
4545 | |
4546 | switch (sa->cmd) { | |
4547 | case BTRFS_QUOTA_CTL_ENABLE: | |
340f1aa2 | 4548 | ret = btrfs_quota_enable(fs_info); |
5d13a37b AJ |
4549 | break; |
4550 | case BTRFS_QUOTA_CTL_DISABLE: | |
340f1aa2 | 4551 | ret = btrfs_quota_disable(fs_info); |
5d13a37b | 4552 | break; |
5d13a37b AJ |
4553 | default: |
4554 | ret = -EINVAL; | |
4555 | break; | |
4556 | } | |
4557 | ||
5d13a37b | 4558 | kfree(sa); |
0b246afa | 4559 | up_write(&fs_info->subvol_sem); |
905b0dda MX |
4560 | drop_write: |
4561 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4562 | return ret; |
4563 | } | |
4564 | ||
905b0dda | 4565 | static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) |
5d13a37b | 4566 | { |
0b246afa JM |
4567 | struct inode *inode = file_inode(file); |
4568 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
4569 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
5d13a37b AJ |
4570 | struct btrfs_ioctl_qgroup_assign_args *sa; |
4571 | struct btrfs_trans_handle *trans; | |
4572 | int ret; | |
4573 | int err; | |
4574 | ||
4575 | if (!capable(CAP_SYS_ADMIN)) | |
4576 | return -EPERM; | |
4577 | ||
905b0dda MX |
4578 | ret = mnt_want_write_file(file); |
4579 | if (ret) | |
4580 | return ret; | |
5d13a37b AJ |
4581 | |
4582 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4583 | if (IS_ERR(sa)) { |
4584 | ret = PTR_ERR(sa); | |
4585 | goto drop_write; | |
4586 | } | |
5d13a37b AJ |
4587 | |
4588 | trans = btrfs_join_transaction(root); | |
4589 | if (IS_ERR(trans)) { | |
4590 | ret = PTR_ERR(trans); | |
4591 | goto out; | |
4592 | } | |
4593 | ||
5d13a37b | 4594 | if (sa->assign) { |
9f8a6ce6 | 4595 | ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst); |
5d13a37b | 4596 | } else { |
39616c27 | 4597 | ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst); |
5d13a37b AJ |
4598 | } |
4599 | ||
e082f563 | 4600 | /* update qgroup status and info */ |
280f8bd2 | 4601 | err = btrfs_run_qgroups(trans); |
e082f563 | 4602 | if (err < 0) |
0b246afa JM |
4603 | btrfs_handle_fs_error(fs_info, err, |
4604 | "failed to update qgroup status and info"); | |
3a45bb20 | 4605 | err = btrfs_end_transaction(trans); |
5d13a37b AJ |
4606 | if (err && !ret) |
4607 | ret = err; | |
4608 | ||
4609 | out: | |
4610 | kfree(sa); | |
905b0dda MX |
4611 | drop_write: |
4612 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4613 | return ret; |
4614 | } | |
4615 | ||
905b0dda | 4616 | static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) |
5d13a37b | 4617 | { |
0b246afa | 4618 | struct inode *inode = file_inode(file); |
0b246afa | 4619 | struct btrfs_root *root = BTRFS_I(inode)->root; |
5d13a37b AJ |
4620 | struct btrfs_ioctl_qgroup_create_args *sa; |
4621 | struct btrfs_trans_handle *trans; | |
4622 | int ret; | |
4623 | int err; | |
4624 | ||
4625 | if (!capable(CAP_SYS_ADMIN)) | |
4626 | return -EPERM; | |
4627 | ||
905b0dda MX |
4628 | ret = mnt_want_write_file(file); |
4629 | if (ret) | |
4630 | return ret; | |
5d13a37b AJ |
4631 | |
4632 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4633 | if (IS_ERR(sa)) { |
4634 | ret = PTR_ERR(sa); | |
4635 | goto drop_write; | |
4636 | } | |
5d13a37b | 4637 | |
d86e56cf MX |
4638 | if (!sa->qgroupid) { |
4639 | ret = -EINVAL; | |
4640 | goto out; | |
4641 | } | |
4642 | ||
5d13a37b AJ |
4643 | trans = btrfs_join_transaction(root); |
4644 | if (IS_ERR(trans)) { | |
4645 | ret = PTR_ERR(trans); | |
4646 | goto out; | |
4647 | } | |
4648 | ||
5d13a37b | 4649 | if (sa->create) { |
49a05ecd | 4650 | ret = btrfs_create_qgroup(trans, sa->qgroupid); |
5d13a37b | 4651 | } else { |
3efbee1d | 4652 | ret = btrfs_remove_qgroup(trans, sa->qgroupid); |
5d13a37b AJ |
4653 | } |
4654 | ||
3a45bb20 | 4655 | err = btrfs_end_transaction(trans); |
5d13a37b AJ |
4656 | if (err && !ret) |
4657 | ret = err; | |
4658 | ||
4659 | out: | |
4660 | kfree(sa); | |
905b0dda MX |
4661 | drop_write: |
4662 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4663 | return ret; |
4664 | } | |
4665 | ||
905b0dda | 4666 | static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) |
5d13a37b | 4667 | { |
0b246afa | 4668 | struct inode *inode = file_inode(file); |
0b246afa | 4669 | struct btrfs_root *root = BTRFS_I(inode)->root; |
5d13a37b AJ |
4670 | struct btrfs_ioctl_qgroup_limit_args *sa; |
4671 | struct btrfs_trans_handle *trans; | |
4672 | int ret; | |
4673 | int err; | |
4674 | u64 qgroupid; | |
4675 | ||
4676 | if (!capable(CAP_SYS_ADMIN)) | |
4677 | return -EPERM; | |
4678 | ||
905b0dda MX |
4679 | ret = mnt_want_write_file(file); |
4680 | if (ret) | |
4681 | return ret; | |
5d13a37b AJ |
4682 | |
4683 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4684 | if (IS_ERR(sa)) { |
4685 | ret = PTR_ERR(sa); | |
4686 | goto drop_write; | |
4687 | } | |
5d13a37b AJ |
4688 | |
4689 | trans = btrfs_join_transaction(root); | |
4690 | if (IS_ERR(trans)) { | |
4691 | ret = PTR_ERR(trans); | |
4692 | goto out; | |
4693 | } | |
4694 | ||
4695 | qgroupid = sa->qgroupid; | |
4696 | if (!qgroupid) { | |
4697 | /* take the current subvol as qgroup */ | |
4698 | qgroupid = root->root_key.objectid; | |
4699 | } | |
4700 | ||
f0042d5e | 4701 | ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim); |
5d13a37b | 4702 | |
3a45bb20 | 4703 | err = btrfs_end_transaction(trans); |
5d13a37b AJ |
4704 | if (err && !ret) |
4705 | ret = err; | |
4706 | ||
4707 | out: | |
4708 | kfree(sa); | |
905b0dda MX |
4709 | drop_write: |
4710 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4711 | return ret; |
4712 | } | |
4713 | ||
2f232036 JS |
4714 | static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) |
4715 | { | |
0b246afa JM |
4716 | struct inode *inode = file_inode(file); |
4717 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
2f232036 JS |
4718 | struct btrfs_ioctl_quota_rescan_args *qsa; |
4719 | int ret; | |
4720 | ||
4721 | if (!capable(CAP_SYS_ADMIN)) | |
4722 | return -EPERM; | |
4723 | ||
4724 | ret = mnt_want_write_file(file); | |
4725 | if (ret) | |
4726 | return ret; | |
4727 | ||
4728 | qsa = memdup_user(arg, sizeof(*qsa)); | |
4729 | if (IS_ERR(qsa)) { | |
4730 | ret = PTR_ERR(qsa); | |
4731 | goto drop_write; | |
4732 | } | |
4733 | ||
4734 | if (qsa->flags) { | |
4735 | ret = -EINVAL; | |
4736 | goto out; | |
4737 | } | |
4738 | ||
0b246afa | 4739 | ret = btrfs_qgroup_rescan(fs_info); |
2f232036 JS |
4740 | |
4741 | out: | |
4742 | kfree(qsa); | |
4743 | drop_write: | |
4744 | mnt_drop_write_file(file); | |
4745 | return ret; | |
4746 | } | |
4747 | ||
b929c1d8 MPS |
4748 | static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info, |
4749 | void __user *arg) | |
2f232036 | 4750 | { |
0afb603a | 4751 | struct btrfs_ioctl_quota_rescan_args qsa = {0}; |
2f232036 JS |
4752 | |
4753 | if (!capable(CAP_SYS_ADMIN)) | |
4754 | return -EPERM; | |
4755 | ||
0b246afa | 4756 | if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { |
0afb603a GR |
4757 | qsa.flags = 1; |
4758 | qsa.progress = fs_info->qgroup_rescan_progress.objectid; | |
2f232036 JS |
4759 | } |
4760 | ||
0afb603a | 4761 | if (copy_to_user(arg, &qsa, sizeof(qsa))) |
991a3dae | 4762 | return -EFAULT; |
2f232036 | 4763 | |
991a3dae | 4764 | return 0; |
2f232036 JS |
4765 | } |
4766 | ||
b929c1d8 MPS |
4767 | static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info, |
4768 | void __user *arg) | |
57254b6e | 4769 | { |
57254b6e JS |
4770 | if (!capable(CAP_SYS_ADMIN)) |
4771 | return -EPERM; | |
4772 | ||
0b246afa | 4773 | return btrfs_qgroup_wait_for_completion(fs_info, true); |
57254b6e JS |
4774 | } |
4775 | ||
abccd00f | 4776 | static long _btrfs_ioctl_set_received_subvol(struct file *file, |
e4fed17a | 4777 | struct user_namespace *mnt_userns, |
abccd00f | 4778 | struct btrfs_ioctl_received_subvol_args *sa) |
8ea05e3a | 4779 | { |
496ad9aa | 4780 | struct inode *inode = file_inode(file); |
0b246afa | 4781 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
8ea05e3a AB |
4782 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4783 | struct btrfs_root_item *root_item = &root->root_item; | |
4784 | struct btrfs_trans_handle *trans; | |
95582b00 | 4785 | struct timespec64 ct = current_time(inode); |
8ea05e3a | 4786 | int ret = 0; |
dd5f9615 | 4787 | int received_uuid_changed; |
8ea05e3a | 4788 | |
e4fed17a | 4789 | if (!inode_owner_or_capable(mnt_userns, inode)) |
bd60ea0f DS |
4790 | return -EPERM; |
4791 | ||
8ea05e3a AB |
4792 | ret = mnt_want_write_file(file); |
4793 | if (ret < 0) | |
4794 | return ret; | |
4795 | ||
0b246afa | 4796 | down_write(&fs_info->subvol_sem); |
8ea05e3a | 4797 | |
4a0cc7ca | 4798 | if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { |
8ea05e3a AB |
4799 | ret = -EINVAL; |
4800 | goto out; | |
4801 | } | |
4802 | ||
4803 | if (btrfs_root_readonly(root)) { | |
4804 | ret = -EROFS; | |
4805 | goto out; | |
4806 | } | |
4807 | ||
dd5f9615 SB |
4808 | /* |
4809 | * 1 - root item | |
4810 | * 2 - uuid items (received uuid + subvol uuid) | |
4811 | */ | |
4812 | trans = btrfs_start_transaction(root, 3); | |
8ea05e3a AB |
4813 | if (IS_ERR(trans)) { |
4814 | ret = PTR_ERR(trans); | |
4815 | trans = NULL; | |
4816 | goto out; | |
4817 | } | |
4818 | ||
4819 | sa->rtransid = trans->transid; | |
4820 | sa->rtime.sec = ct.tv_sec; | |
4821 | sa->rtime.nsec = ct.tv_nsec; | |
4822 | ||
dd5f9615 SB |
4823 | received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, |
4824 | BTRFS_UUID_SIZE); | |
4825 | if (received_uuid_changed && | |
d87ff758 | 4826 | !btrfs_is_empty_uuid(root_item->received_uuid)) { |
d1957791 | 4827 | ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, |
d87ff758 NB |
4828 | BTRFS_UUID_KEY_RECEIVED_SUBVOL, |
4829 | root->root_key.objectid); | |
4830 | if (ret && ret != -ENOENT) { | |
4831 | btrfs_abort_transaction(trans, ret); | |
4832 | btrfs_end_transaction(trans); | |
4833 | goto out; | |
4834 | } | |
4835 | } | |
8ea05e3a AB |
4836 | memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); |
4837 | btrfs_set_root_stransid(root_item, sa->stransid); | |
4838 | btrfs_set_root_rtransid(root_item, sa->rtransid); | |
3cae210f QW |
4839 | btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); |
4840 | btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); | |
4841 | btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); | |
4842 | btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); | |
8ea05e3a | 4843 | |
0b246afa | 4844 | ret = btrfs_update_root(trans, fs_info->tree_root, |
8ea05e3a AB |
4845 | &root->root_key, &root->root_item); |
4846 | if (ret < 0) { | |
3a45bb20 | 4847 | btrfs_end_transaction(trans); |
8ea05e3a | 4848 | goto out; |
dd5f9615 SB |
4849 | } |
4850 | if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { | |
cdb345a8 | 4851 | ret = btrfs_uuid_tree_add(trans, sa->uuid, |
dd5f9615 SB |
4852 | BTRFS_UUID_KEY_RECEIVED_SUBVOL, |
4853 | root->root_key.objectid); | |
4854 | if (ret < 0 && ret != -EEXIST) { | |
66642832 | 4855 | btrfs_abort_transaction(trans, ret); |
efd38150 | 4856 | btrfs_end_transaction(trans); |
8ea05e3a | 4857 | goto out; |
dd5f9615 SB |
4858 | } |
4859 | } | |
3a45bb20 | 4860 | ret = btrfs_commit_transaction(trans); |
abccd00f | 4861 | out: |
0b246afa | 4862 | up_write(&fs_info->subvol_sem); |
abccd00f HM |
4863 | mnt_drop_write_file(file); |
4864 | return ret; | |
4865 | } | |
4866 | ||
4867 | #ifdef CONFIG_64BIT | |
4868 | static long btrfs_ioctl_set_received_subvol_32(struct file *file, | |
4869 | void __user *arg) | |
4870 | { | |
4871 | struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; | |
4872 | struct btrfs_ioctl_received_subvol_args *args64 = NULL; | |
4873 | int ret = 0; | |
4874 | ||
4875 | args32 = memdup_user(arg, sizeof(*args32)); | |
7b9ea627 SV |
4876 | if (IS_ERR(args32)) |
4877 | return PTR_ERR(args32); | |
abccd00f | 4878 | |
8d2db785 | 4879 | args64 = kmalloc(sizeof(*args64), GFP_KERNEL); |
84dbeb87 DC |
4880 | if (!args64) { |
4881 | ret = -ENOMEM; | |
abccd00f HM |
4882 | goto out; |
4883 | } | |
4884 | ||
4885 | memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); | |
4886 | args64->stransid = args32->stransid; | |
4887 | args64->rtransid = args32->rtransid; | |
4888 | args64->stime.sec = args32->stime.sec; | |
4889 | args64->stime.nsec = args32->stime.nsec; | |
4890 | args64->rtime.sec = args32->rtime.sec; | |
4891 | args64->rtime.nsec = args32->rtime.nsec; | |
4892 | args64->flags = args32->flags; | |
4893 | ||
e4fed17a | 4894 | ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), args64); |
abccd00f HM |
4895 | if (ret) |
4896 | goto out; | |
4897 | ||
4898 | memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); | |
4899 | args32->stransid = args64->stransid; | |
4900 | args32->rtransid = args64->rtransid; | |
4901 | args32->stime.sec = args64->stime.sec; | |
4902 | args32->stime.nsec = args64->stime.nsec; | |
4903 | args32->rtime.sec = args64->rtime.sec; | |
4904 | args32->rtime.nsec = args64->rtime.nsec; | |
4905 | args32->flags = args64->flags; | |
4906 | ||
4907 | ret = copy_to_user(arg, args32, sizeof(*args32)); | |
4908 | if (ret) | |
4909 | ret = -EFAULT; | |
4910 | ||
4911 | out: | |
4912 | kfree(args32); | |
4913 | kfree(args64); | |
4914 | return ret; | |
4915 | } | |
4916 | #endif | |
4917 | ||
4918 | static long btrfs_ioctl_set_received_subvol(struct file *file, | |
4919 | void __user *arg) | |
4920 | { | |
4921 | struct btrfs_ioctl_received_subvol_args *sa = NULL; | |
4922 | int ret = 0; | |
4923 | ||
4924 | sa = memdup_user(arg, sizeof(*sa)); | |
7b9ea627 SV |
4925 | if (IS_ERR(sa)) |
4926 | return PTR_ERR(sa); | |
abccd00f | 4927 | |
e4fed17a | 4928 | ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), sa); |
abccd00f HM |
4929 | |
4930 | if (ret) | |
4931 | goto out; | |
4932 | ||
8ea05e3a AB |
4933 | ret = copy_to_user(arg, sa, sizeof(*sa)); |
4934 | if (ret) | |
4935 | ret = -EFAULT; | |
4936 | ||
4937 | out: | |
4938 | kfree(sa); | |
8ea05e3a AB |
4939 | return ret; |
4940 | } | |
4941 | ||
b929c1d8 MPS |
4942 | static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info, |
4943 | void __user *arg) | |
867ab667 | 4944 | { |
a1b83ac5 | 4945 | size_t len; |
867ab667 | 4946 | int ret; |
a1b83ac5 AJ |
4947 | char label[BTRFS_LABEL_SIZE]; |
4948 | ||
0b246afa JM |
4949 | spin_lock(&fs_info->super_lock); |
4950 | memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); | |
4951 | spin_unlock(&fs_info->super_lock); | |
a1b83ac5 AJ |
4952 | |
4953 | len = strnlen(label, BTRFS_LABEL_SIZE); | |
867ab667 | 4954 | |
4955 | if (len == BTRFS_LABEL_SIZE) { | |
0b246afa JM |
4956 | btrfs_warn(fs_info, |
4957 | "label is too long, return the first %zu bytes", | |
4958 | --len); | |
867ab667 | 4959 | } |
4960 | ||
867ab667 | 4961 | ret = copy_to_user(arg, label, len); |
867ab667 | 4962 | |
4963 | return ret ? -EFAULT : 0; | |
4964 | } | |
4965 | ||
a8bfd4ab | 4966 | static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) |
4967 | { | |
0b246afa JM |
4968 | struct inode *inode = file_inode(file); |
4969 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
4970 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
4971 | struct btrfs_super_block *super_block = fs_info->super_copy; | |
a8bfd4ab | 4972 | struct btrfs_trans_handle *trans; |
4973 | char label[BTRFS_LABEL_SIZE]; | |
4974 | int ret; | |
4975 | ||
4976 | if (!capable(CAP_SYS_ADMIN)) | |
4977 | return -EPERM; | |
4978 | ||
4979 | if (copy_from_user(label, arg, sizeof(label))) | |
4980 | return -EFAULT; | |
4981 | ||
4982 | if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { | |
0b246afa | 4983 | btrfs_err(fs_info, |
5d163e0e JM |
4984 | "unable to set label with more than %d bytes", |
4985 | BTRFS_LABEL_SIZE - 1); | |
a8bfd4ab | 4986 | return -EINVAL; |
4987 | } | |
4988 | ||
4989 | ret = mnt_want_write_file(file); | |
4990 | if (ret) | |
4991 | return ret; | |
4992 | ||
a8bfd4ab | 4993 | trans = btrfs_start_transaction(root, 0); |
4994 | if (IS_ERR(trans)) { | |
4995 | ret = PTR_ERR(trans); | |
4996 | goto out_unlock; | |
4997 | } | |
4998 | ||
0b246afa | 4999 | spin_lock(&fs_info->super_lock); |
a8bfd4ab | 5000 | strcpy(super_block->label, label); |
0b246afa | 5001 | spin_unlock(&fs_info->super_lock); |
3a45bb20 | 5002 | ret = btrfs_commit_transaction(trans); |
a8bfd4ab | 5003 | |
5004 | out_unlock: | |
a8bfd4ab | 5005 | mnt_drop_write_file(file); |
5006 | return ret; | |
5007 | } | |
5008 | ||
2eaa055f JM |
5009 | #define INIT_FEATURE_FLAGS(suffix) \ |
5010 | { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ | |
5011 | .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ | |
5012 | .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } | |
5013 | ||
d5131b65 | 5014 | int btrfs_ioctl_get_supported_features(void __user *arg) |
2eaa055f | 5015 | { |
4d4ab6d6 | 5016 | static const struct btrfs_ioctl_feature_flags features[3] = { |
2eaa055f JM |
5017 | INIT_FEATURE_FLAGS(SUPP), |
5018 | INIT_FEATURE_FLAGS(SAFE_SET), | |
5019 | INIT_FEATURE_FLAGS(SAFE_CLEAR) | |
5020 | }; | |
5021 | ||
5022 | if (copy_to_user(arg, &features, sizeof(features))) | |
5023 | return -EFAULT; | |
5024 | ||
5025 | return 0; | |
5026 | } | |
5027 | ||
b929c1d8 MPS |
5028 | static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info, |
5029 | void __user *arg) | |
2eaa055f | 5030 | { |
0b246afa | 5031 | struct btrfs_super_block *super_block = fs_info->super_copy; |
2eaa055f JM |
5032 | struct btrfs_ioctl_feature_flags features; |
5033 | ||
5034 | features.compat_flags = btrfs_super_compat_flags(super_block); | |
5035 | features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); | |
5036 | features.incompat_flags = btrfs_super_incompat_flags(super_block); | |
5037 | ||
5038 | if (copy_to_user(arg, &features, sizeof(features))) | |
5039 | return -EFAULT; | |
5040 | ||
5041 | return 0; | |
5042 | } | |
5043 | ||
2ff7e61e | 5044 | static int check_feature_bits(struct btrfs_fs_info *fs_info, |
3b02a68a | 5045 | enum btrfs_feature_set set, |
2eaa055f JM |
5046 | u64 change_mask, u64 flags, u64 supported_flags, |
5047 | u64 safe_set, u64 safe_clear) | |
5048 | { | |
f10152bc | 5049 | const char *type = btrfs_feature_set_name(set); |
3b02a68a | 5050 | char *names; |
2eaa055f JM |
5051 | u64 disallowed, unsupported; |
5052 | u64 set_mask = flags & change_mask; | |
5053 | u64 clear_mask = ~flags & change_mask; | |
5054 | ||
5055 | unsupported = set_mask & ~supported_flags; | |
5056 | if (unsupported) { | |
3b02a68a JM |
5057 | names = btrfs_printable_features(set, unsupported); |
5058 | if (names) { | |
0b246afa JM |
5059 | btrfs_warn(fs_info, |
5060 | "this kernel does not support the %s feature bit%s", | |
5061 | names, strchr(names, ',') ? "s" : ""); | |
3b02a68a JM |
5062 | kfree(names); |
5063 | } else | |
0b246afa JM |
5064 | btrfs_warn(fs_info, |
5065 | "this kernel does not support %s bits 0x%llx", | |
5066 | type, unsupported); | |
2eaa055f JM |
5067 | return -EOPNOTSUPP; |
5068 | } | |
5069 | ||
5070 | disallowed = set_mask & ~safe_set; | |
5071 | if (disallowed) { | |
3b02a68a JM |
5072 | names = btrfs_printable_features(set, disallowed); |
5073 | if (names) { | |
0b246afa JM |
5074 | btrfs_warn(fs_info, |
5075 | "can't set the %s feature bit%s while mounted", | |
5076 | names, strchr(names, ',') ? "s" : ""); | |
3b02a68a JM |
5077 | kfree(names); |
5078 | } else | |
0b246afa JM |
5079 | btrfs_warn(fs_info, |
5080 | "can't set %s bits 0x%llx while mounted", | |
5081 | type, disallowed); | |
2eaa055f JM |
5082 | return -EPERM; |
5083 | } | |
5084 | ||
5085 | disallowed = clear_mask & ~safe_clear; | |
5086 | if (disallowed) { | |
3b02a68a JM |
5087 | names = btrfs_printable_features(set, disallowed); |
5088 | if (names) { | |
0b246afa JM |
5089 | btrfs_warn(fs_info, |
5090 | "can't clear the %s feature bit%s while mounted", | |
5091 | names, strchr(names, ',') ? "s" : ""); | |
3b02a68a JM |
5092 | kfree(names); |
5093 | } else | |
0b246afa JM |
5094 | btrfs_warn(fs_info, |
5095 | "can't clear %s bits 0x%llx while mounted", | |
5096 | type, disallowed); | |
2eaa055f JM |
5097 | return -EPERM; |
5098 | } | |
5099 | ||
5100 | return 0; | |
5101 | } | |
5102 | ||
2ff7e61e JM |
5103 | #define check_feature(fs_info, change_mask, flags, mask_base) \ |
5104 | check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ | |
2eaa055f JM |
5105 | BTRFS_FEATURE_ ## mask_base ## _SUPP, \ |
5106 | BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ | |
5107 | BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) | |
5108 | ||
5109 | static int btrfs_ioctl_set_features(struct file *file, void __user *arg) | |
5110 | { | |
0b246afa JM |
5111 | struct inode *inode = file_inode(file); |
5112 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
5113 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
5114 | struct btrfs_super_block *super_block = fs_info->super_copy; | |
2eaa055f JM |
5115 | struct btrfs_ioctl_feature_flags flags[2]; |
5116 | struct btrfs_trans_handle *trans; | |
5117 | u64 newflags; | |
5118 | int ret; | |
5119 | ||
5120 | if (!capable(CAP_SYS_ADMIN)) | |
5121 | return -EPERM; | |
5122 | ||
5123 | if (copy_from_user(flags, arg, sizeof(flags))) | |
5124 | return -EFAULT; | |
5125 | ||
5126 | /* Nothing to do */ | |
5127 | if (!flags[0].compat_flags && !flags[0].compat_ro_flags && | |
5128 | !flags[0].incompat_flags) | |
5129 | return 0; | |
5130 | ||
2ff7e61e | 5131 | ret = check_feature(fs_info, flags[0].compat_flags, |
2eaa055f JM |
5132 | flags[1].compat_flags, COMPAT); |
5133 | if (ret) | |
5134 | return ret; | |
5135 | ||
2ff7e61e | 5136 | ret = check_feature(fs_info, flags[0].compat_ro_flags, |
2eaa055f JM |
5137 | flags[1].compat_ro_flags, COMPAT_RO); |
5138 | if (ret) | |
5139 | return ret; | |
5140 | ||
2ff7e61e | 5141 | ret = check_feature(fs_info, flags[0].incompat_flags, |
2eaa055f JM |
5142 | flags[1].incompat_flags, INCOMPAT); |
5143 | if (ret) | |
5144 | return ret; | |
5145 | ||
7ab19625 DS |
5146 | ret = mnt_want_write_file(file); |
5147 | if (ret) | |
5148 | return ret; | |
5149 | ||
8051aa1a | 5150 | trans = btrfs_start_transaction(root, 0); |
7ab19625 DS |
5151 | if (IS_ERR(trans)) { |
5152 | ret = PTR_ERR(trans); | |
5153 | goto out_drop_write; | |
5154 | } | |
2eaa055f | 5155 | |
0b246afa | 5156 | spin_lock(&fs_info->super_lock); |
2eaa055f JM |
5157 | newflags = btrfs_super_compat_flags(super_block); |
5158 | newflags |= flags[0].compat_flags & flags[1].compat_flags; | |
5159 | newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); | |
5160 | btrfs_set_super_compat_flags(super_block, newflags); | |
5161 | ||
5162 | newflags = btrfs_super_compat_ro_flags(super_block); | |
5163 | newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; | |
5164 | newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); | |
5165 | btrfs_set_super_compat_ro_flags(super_block, newflags); | |
5166 | ||
5167 | newflags = btrfs_super_incompat_flags(super_block); | |
5168 | newflags |= flags[0].incompat_flags & flags[1].incompat_flags; | |
5169 | newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); | |
5170 | btrfs_set_super_incompat_flags(super_block, newflags); | |
0b246afa | 5171 | spin_unlock(&fs_info->super_lock); |
2eaa055f | 5172 | |
3a45bb20 | 5173 | ret = btrfs_commit_transaction(trans); |
7ab19625 DS |
5174 | out_drop_write: |
5175 | mnt_drop_write_file(file); | |
5176 | ||
5177 | return ret; | |
2eaa055f JM |
5178 | } |
5179 | ||
9ad12305 | 5180 | static int _btrfs_ioctl_send(struct inode *inode, void __user *argp, bool compat) |
2351f431 JB |
5181 | { |
5182 | struct btrfs_ioctl_send_args *arg; | |
5183 | int ret; | |
5184 | ||
5185 | if (compat) { | |
5186 | #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) | |
5187 | struct btrfs_ioctl_send_args_32 args32; | |
5188 | ||
5189 | ret = copy_from_user(&args32, argp, sizeof(args32)); | |
5190 | if (ret) | |
5191 | return -EFAULT; | |
5192 | arg = kzalloc(sizeof(*arg), GFP_KERNEL); | |
5193 | if (!arg) | |
5194 | return -ENOMEM; | |
5195 | arg->send_fd = args32.send_fd; | |
5196 | arg->clone_sources_count = args32.clone_sources_count; | |
5197 | arg->clone_sources = compat_ptr(args32.clone_sources); | |
5198 | arg->parent_root = args32.parent_root; | |
5199 | arg->flags = args32.flags; | |
5200 | memcpy(arg->reserved, args32.reserved, | |
5201 | sizeof(args32.reserved)); | |
5202 | #else | |
5203 | return -ENOTTY; | |
5204 | #endif | |
5205 | } else { | |
5206 | arg = memdup_user(argp, sizeof(*arg)); | |
5207 | if (IS_ERR(arg)) | |
5208 | return PTR_ERR(arg); | |
5209 | } | |
9ad12305 | 5210 | ret = btrfs_ioctl_send(inode, arg); |
2351f431 JB |
5211 | kfree(arg); |
5212 | return ret; | |
5213 | } | |
5214 | ||
1881fba8 OS |
5215 | static int btrfs_ioctl_encoded_read(struct file *file, void __user *argp, |
5216 | bool compat) | |
5217 | { | |
5218 | struct btrfs_ioctl_encoded_io_args args = { 0 }; | |
5219 | size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args, | |
5220 | flags); | |
5221 | size_t copy_end; | |
5222 | struct iovec iovstack[UIO_FASTIOV]; | |
5223 | struct iovec *iov = iovstack; | |
5224 | struct iov_iter iter; | |
5225 | loff_t pos; | |
5226 | struct kiocb kiocb; | |
5227 | ssize_t ret; | |
5228 | ||
5229 | if (!capable(CAP_SYS_ADMIN)) { | |
5230 | ret = -EPERM; | |
5231 | goto out_acct; | |
5232 | } | |
5233 | ||
5234 | if (compat) { | |
5235 | #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) | |
5236 | struct btrfs_ioctl_encoded_io_args_32 args32; | |
5237 | ||
5238 | copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32, | |
5239 | flags); | |
5240 | if (copy_from_user(&args32, argp, copy_end)) { | |
5241 | ret = -EFAULT; | |
5242 | goto out_acct; | |
5243 | } | |
5244 | args.iov = compat_ptr(args32.iov); | |
5245 | args.iovcnt = args32.iovcnt; | |
5246 | args.offset = args32.offset; | |
5247 | args.flags = args32.flags; | |
5248 | #else | |
5249 | return -ENOTTY; | |
5250 | #endif | |
5251 | } else { | |
5252 | copy_end = copy_end_kernel; | |
5253 | if (copy_from_user(&args, argp, copy_end)) { | |
5254 | ret = -EFAULT; | |
5255 | goto out_acct; | |
5256 | } | |
5257 | } | |
5258 | if (args.flags != 0) { | |
5259 | ret = -EINVAL; | |
5260 | goto out_acct; | |
5261 | } | |
5262 | ||
5263 | ret = import_iovec(READ, args.iov, args.iovcnt, ARRAY_SIZE(iovstack), | |
5264 | &iov, &iter); | |
5265 | if (ret < 0) | |
5266 | goto out_acct; | |
5267 | ||
5268 | if (iov_iter_count(&iter) == 0) { | |
5269 | ret = 0; | |
5270 | goto out_iov; | |
5271 | } | |
5272 | pos = args.offset; | |
5273 | ret = rw_verify_area(READ, file, &pos, args.len); | |
5274 | if (ret < 0) | |
5275 | goto out_iov; | |
5276 | ||
5277 | init_sync_kiocb(&kiocb, file); | |
5278 | kiocb.ki_pos = pos; | |
5279 | ||
5280 | ret = btrfs_encoded_read(&kiocb, &iter, &args); | |
5281 | if (ret >= 0) { | |
5282 | fsnotify_access(file); | |
5283 | if (copy_to_user(argp + copy_end, | |
5284 | (char *)&args + copy_end_kernel, | |
5285 | sizeof(args) - copy_end_kernel)) | |
5286 | ret = -EFAULT; | |
5287 | } | |
5288 | ||
5289 | out_iov: | |
5290 | kfree(iov); | |
5291 | out_acct: | |
5292 | if (ret > 0) | |
5293 | add_rchar(current, ret); | |
5294 | inc_syscr(current); | |
5295 | return ret; | |
5296 | } | |
5297 | ||
f46b5a66 CH |
5298 | long btrfs_ioctl(struct file *file, unsigned int |
5299 | cmd, unsigned long arg) | |
5300 | { | |
0b246afa JM |
5301 | struct inode *inode = file_inode(file); |
5302 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
5303 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
4bcabaa3 | 5304 | void __user *argp = (void __user *)arg; |
f46b5a66 CH |
5305 | |
5306 | switch (cmd) { | |
6cbff00f | 5307 | case FS_IOC_GETVERSION: |
dc408ccd | 5308 | return btrfs_ioctl_getversion(inode, argp); |
40cf931f | 5309 | case FS_IOC_GETFSLABEL: |
b929c1d8 | 5310 | return btrfs_ioctl_get_fslabel(fs_info, argp); |
40cf931f ES |
5311 | case FS_IOC_SETFSLABEL: |
5312 | return btrfs_ioctl_set_fslabel(file, argp); | |
f7039b1d | 5313 | case FITRIM: |
b929c1d8 | 5314 | return btrfs_ioctl_fitrim(fs_info, argp); |
f46b5a66 | 5315 | case BTRFS_IOC_SNAP_CREATE: |
fa0d2b9b | 5316 | return btrfs_ioctl_snap_create(file, argp, 0); |
fdfb1e4f | 5317 | case BTRFS_IOC_SNAP_CREATE_V2: |
fa0d2b9b | 5318 | return btrfs_ioctl_snap_create_v2(file, argp, 0); |
3de4586c | 5319 | case BTRFS_IOC_SUBVOL_CREATE: |
fa0d2b9b | 5320 | return btrfs_ioctl_snap_create(file, argp, 1); |
6f72c7e2 AJ |
5321 | case BTRFS_IOC_SUBVOL_CREATE_V2: |
5322 | return btrfs_ioctl_snap_create_v2(file, argp, 1); | |
76dda93c | 5323 | case BTRFS_IOC_SNAP_DESTROY: |
949964c9 MPS |
5324 | return btrfs_ioctl_snap_destroy(file, argp, false); |
5325 | case BTRFS_IOC_SNAP_DESTROY_V2: | |
5326 | return btrfs_ioctl_snap_destroy(file, argp, true); | |
0caa102d | 5327 | case BTRFS_IOC_SUBVOL_GETFLAGS: |
9ad12305 | 5328 | return btrfs_ioctl_subvol_getflags(inode, argp); |
0caa102d LZ |
5329 | case BTRFS_IOC_SUBVOL_SETFLAGS: |
5330 | return btrfs_ioctl_subvol_setflags(file, argp); | |
6ef5ed0d JB |
5331 | case BTRFS_IOC_DEFAULT_SUBVOL: |
5332 | return btrfs_ioctl_default_subvol(file, argp); | |
f46b5a66 | 5333 | case BTRFS_IOC_DEFRAG: |
1e701a32 CM |
5334 | return btrfs_ioctl_defrag(file, NULL); |
5335 | case BTRFS_IOC_DEFRAG_RANGE: | |
5336 | return btrfs_ioctl_defrag(file, argp); | |
f46b5a66 | 5337 | case BTRFS_IOC_RESIZE: |
198605a8 | 5338 | return btrfs_ioctl_resize(file, argp); |
f46b5a66 | 5339 | case BTRFS_IOC_ADD_DEV: |
2ff7e61e | 5340 | return btrfs_ioctl_add_dev(fs_info, argp); |
f46b5a66 | 5341 | case BTRFS_IOC_RM_DEV: |
da24927b | 5342 | return btrfs_ioctl_rm_dev(file, argp); |
6b526ed7 AJ |
5343 | case BTRFS_IOC_RM_DEV_V2: |
5344 | return btrfs_ioctl_rm_dev_v2(file, argp); | |
475f6387 | 5345 | case BTRFS_IOC_FS_INFO: |
2ff7e61e | 5346 | return btrfs_ioctl_fs_info(fs_info, argp); |
475f6387 | 5347 | case BTRFS_IOC_DEV_INFO: |
2ff7e61e | 5348 | return btrfs_ioctl_dev_info(fs_info, argp); |
f46b5a66 | 5349 | case BTRFS_IOC_BALANCE: |
9ba1f6e4 | 5350 | return btrfs_ioctl_balance(file, NULL); |
ac8e9819 | 5351 | case BTRFS_IOC_TREE_SEARCH: |
9ad12305 | 5352 | return btrfs_ioctl_tree_search(inode, argp); |
cc68a8a5 | 5353 | case BTRFS_IOC_TREE_SEARCH_V2: |
9ad12305 | 5354 | return btrfs_ioctl_tree_search_v2(inode, argp); |
ac8e9819 | 5355 | case BTRFS_IOC_INO_LOOKUP: |
dc408ccd | 5356 | return btrfs_ioctl_ino_lookup(root, argp); |
d7728c96 JS |
5357 | case BTRFS_IOC_INO_PATHS: |
5358 | return btrfs_ioctl_ino_to_path(root, argp); | |
5359 | case BTRFS_IOC_LOGICAL_INO: | |
d24a67b2 ZB |
5360 | return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); |
5361 | case BTRFS_IOC_LOGICAL_INO_V2: | |
5362 | return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); | |
1406e432 | 5363 | case BTRFS_IOC_SPACE_INFO: |
2ff7e61e | 5364 | return btrfs_ioctl_space_info(fs_info, argp); |
9b199859 FDBM |
5365 | case BTRFS_IOC_SYNC: { |
5366 | int ret; | |
5367 | ||
9db4dc24 | 5368 | ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false); |
9b199859 FDBM |
5369 | if (ret) |
5370 | return ret; | |
0b246afa | 5371 | ret = btrfs_sync_fs(inode->i_sb, 1); |
2fad4e83 DS |
5372 | /* |
5373 | * The transaction thread may want to do more work, | |
01327610 | 5374 | * namely it pokes the cleaner kthread that will start |
2fad4e83 DS |
5375 | * processing uncleaned subvols. |
5376 | */ | |
0b246afa | 5377 | wake_up_process(fs_info->transaction_kthread); |
9b199859 FDBM |
5378 | return ret; |
5379 | } | |
46204592 | 5380 | case BTRFS_IOC_START_SYNC: |
9a8c28be | 5381 | return btrfs_ioctl_start_sync(root, argp); |
46204592 | 5382 | case BTRFS_IOC_WAIT_SYNC: |
2ff7e61e | 5383 | return btrfs_ioctl_wait_sync(fs_info, argp); |
475f6387 | 5384 | case BTRFS_IOC_SCRUB: |
b8e95489 | 5385 | return btrfs_ioctl_scrub(file, argp); |
475f6387 | 5386 | case BTRFS_IOC_SCRUB_CANCEL: |
2ff7e61e | 5387 | return btrfs_ioctl_scrub_cancel(fs_info); |
475f6387 | 5388 | case BTRFS_IOC_SCRUB_PROGRESS: |
2ff7e61e | 5389 | return btrfs_ioctl_scrub_progress(fs_info, argp); |
c9e9f97b | 5390 | case BTRFS_IOC_BALANCE_V2: |
9ba1f6e4 | 5391 | return btrfs_ioctl_balance(file, argp); |
837d5b6e | 5392 | case BTRFS_IOC_BALANCE_CTL: |
2ff7e61e | 5393 | return btrfs_ioctl_balance_ctl(fs_info, arg); |
19a39dce | 5394 | case BTRFS_IOC_BALANCE_PROGRESS: |
2ff7e61e | 5395 | return btrfs_ioctl_balance_progress(fs_info, argp); |
8ea05e3a AB |
5396 | case BTRFS_IOC_SET_RECEIVED_SUBVOL: |
5397 | return btrfs_ioctl_set_received_subvol(file, argp); | |
abccd00f HM |
5398 | #ifdef CONFIG_64BIT |
5399 | case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: | |
5400 | return btrfs_ioctl_set_received_subvol_32(file, argp); | |
5401 | #endif | |
31db9f7c | 5402 | case BTRFS_IOC_SEND: |
9ad12305 | 5403 | return _btrfs_ioctl_send(inode, argp, false); |
2351f431 JB |
5404 | #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) |
5405 | case BTRFS_IOC_SEND_32: | |
9ad12305 | 5406 | return _btrfs_ioctl_send(inode, argp, true); |
2351f431 | 5407 | #endif |
c11d2c23 | 5408 | case BTRFS_IOC_GET_DEV_STATS: |
2ff7e61e | 5409 | return btrfs_ioctl_get_dev_stats(fs_info, argp); |
5d13a37b | 5410 | case BTRFS_IOC_QUOTA_CTL: |
905b0dda | 5411 | return btrfs_ioctl_quota_ctl(file, argp); |
5d13a37b | 5412 | case BTRFS_IOC_QGROUP_ASSIGN: |
905b0dda | 5413 | return btrfs_ioctl_qgroup_assign(file, argp); |
5d13a37b | 5414 | case BTRFS_IOC_QGROUP_CREATE: |
905b0dda | 5415 | return btrfs_ioctl_qgroup_create(file, argp); |
5d13a37b | 5416 | case BTRFS_IOC_QGROUP_LIMIT: |
905b0dda | 5417 | return btrfs_ioctl_qgroup_limit(file, argp); |
2f232036 JS |
5418 | case BTRFS_IOC_QUOTA_RESCAN: |
5419 | return btrfs_ioctl_quota_rescan(file, argp); | |
5420 | case BTRFS_IOC_QUOTA_RESCAN_STATUS: | |
b929c1d8 | 5421 | return btrfs_ioctl_quota_rescan_status(fs_info, argp); |
57254b6e | 5422 | case BTRFS_IOC_QUOTA_RESCAN_WAIT: |
b929c1d8 | 5423 | return btrfs_ioctl_quota_rescan_wait(fs_info, argp); |
3f6bcfbd | 5424 | case BTRFS_IOC_DEV_REPLACE: |
2ff7e61e | 5425 | return btrfs_ioctl_dev_replace(fs_info, argp); |
2eaa055f | 5426 | case BTRFS_IOC_GET_SUPPORTED_FEATURES: |
d5131b65 | 5427 | return btrfs_ioctl_get_supported_features(argp); |
2eaa055f | 5428 | case BTRFS_IOC_GET_FEATURES: |
b929c1d8 | 5429 | return btrfs_ioctl_get_features(fs_info, argp); |
2eaa055f JM |
5430 | case BTRFS_IOC_SET_FEATURES: |
5431 | return btrfs_ioctl_set_features(file, argp); | |
b64ec075 | 5432 | case BTRFS_IOC_GET_SUBVOL_INFO: |
9ad12305 | 5433 | return btrfs_ioctl_get_subvol_info(inode, argp); |
42e4b520 | 5434 | case BTRFS_IOC_GET_SUBVOL_ROOTREF: |
dc408ccd | 5435 | return btrfs_ioctl_get_subvol_rootref(root, argp); |
23d0b79d TM |
5436 | case BTRFS_IOC_INO_LOOKUP_USER: |
5437 | return btrfs_ioctl_ino_lookup_user(file, argp); | |
14605409 BB |
5438 | case FS_IOC_ENABLE_VERITY: |
5439 | return fsverity_ioctl_enable(file, (const void __user *)argp); | |
5440 | case FS_IOC_MEASURE_VERITY: | |
5441 | return fsverity_ioctl_measure(file, argp); | |
1881fba8 OS |
5442 | case BTRFS_IOC_ENCODED_READ: |
5443 | return btrfs_ioctl_encoded_read(file, argp, false); | |
5444 | #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) | |
5445 | case BTRFS_IOC_ENCODED_READ_32: | |
5446 | return btrfs_ioctl_encoded_read(file, argp, true); | |
5447 | #endif | |
f46b5a66 CH |
5448 | } |
5449 | ||
5450 | return -ENOTTY; | |
5451 | } | |
4c63c245 LD |
5452 | |
5453 | #ifdef CONFIG_COMPAT | |
5454 | long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) | |
5455 | { | |
2a362249 JM |
5456 | /* |
5457 | * These all access 32-bit values anyway so no further | |
5458 | * handling is necessary. | |
5459 | */ | |
4c63c245 | 5460 | switch (cmd) { |
4c63c245 LD |
5461 | case FS_IOC32_GETVERSION: |
5462 | cmd = FS_IOC_GETVERSION; | |
5463 | break; | |
4c63c245 LD |
5464 | } |
5465 | ||
5466 | return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); | |
5467 | } | |
5468 | #endif |