]>
Commit | Line | Data |
---|---|---|
d2912cb1 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
e9be9d5e MS |
2 | /* |
3 | * | |
4 | * Copyright (C) 2011 Novell Inc. | |
e9be9d5e MS |
5 | */ |
6 | ||
5b825c3a | 7 | #include <uapi/linux/magic.h> |
e9be9d5e MS |
8 | #include <linux/fs.h> |
9 | #include <linux/namei.h> | |
10 | #include <linux/xattr.h> | |
e9be9d5e | 11 | #include <linux/mount.h> |
e9be9d5e MS |
12 | #include <linux/parser.h> |
13 | #include <linux/module.h> | |
cc259639 | 14 | #include <linux/statfs.h> |
f45827e8 | 15 | #include <linux/seq_file.h> |
d837a49b | 16 | #include <linux/posix_acl_xattr.h> |
e487d889 | 17 | #include <linux/exportfs.h> |
2b1a7746 | 18 | #include <linux/file.h> |
1784fbc2 | 19 | #include <linux/fs_context.h> |
dcb399de | 20 | #include <linux/fs_parser.h> |
e9be9d5e | 21 | #include "overlayfs.h" |
7fb7998b | 22 | #include "params.h" |
e9be9d5e MS |
23 | |
24 | MODULE_AUTHOR("Miklos Szeredi <miklos@szeredi.hu>"); | |
25 | MODULE_DESCRIPTION("Overlay filesystem"); | |
26 | MODULE_LICENSE("GPL"); | |
27 | ||
e9be9d5e MS |
28 | |
29 | struct ovl_dir_cache; | |
30 | ||
2d902671 | 31 | static struct dentry *ovl_d_real(struct dentry *dentry, |
fb16043b | 32 | const struct inode *inode) |
d101a125 | 33 | { |
cef4cbff | 34 | struct dentry *real = NULL, *lower; |
184996e9 | 35 | int err; |
d101a125 | 36 | |
def3ae83 AG |
37 | /* |
38 | * vfs is only expected to call d_real() with NULL from d_real_inode() | |
39 | * and with overlay inode from file_dentry() on an overlay file. | |
40 | * | |
41 | * TODO: remove @inode argument from d_real() API, remove code in this | |
42 | * function that deals with non-NULL @inode and remove d_real() call | |
43 | * from file_dentry(). | |
44 | */ | |
e8c985ba MS |
45 | if (inode && d_inode(dentry) == inode) |
46 | return dentry; | |
def3ae83 AG |
47 | else if (inode) |
48 | goto bug; | |
e8c985ba | 49 | |
ca4c8a3a | 50 | if (!d_is_reg(dentry)) { |
def3ae83 AG |
51 | /* d_real_inode() is only relevant for regular files */ |
52 | return dentry; | |
d101a125 MS |
53 | } |
54 | ||
55 | real = ovl_dentry_upper(dentry); | |
2c3d7358 | 56 | if (real && (inode == d_inode(real))) |
d101a125 MS |
57 | return real; |
58 | ||
2c3d7358 VG |
59 | if (real && !inode && ovl_has_upperdata(d_inode(dentry))) |
60 | return real; | |
61 | ||
41665644 | 62 | /* |
42dd69ae | 63 | * Best effort lazy lookup of lowerdata for !inode case to return |
41665644 AG |
64 | * the real lowerdata dentry. The only current caller of d_real() with |
65 | * NULL inode is d_real_inode() from trace_uprobe and this caller is | |
66 | * likely going to be followed reading from the file, before placing | |
67 | * uprobes on offset within the file, so lowerdata should be available | |
68 | * when setting the uprobe. | |
69 | */ | |
184996e9 AL |
70 | err = ovl_verify_lowerdata(dentry); |
71 | if (err) | |
72 | goto bug; | |
cef4cbff MS |
73 | lower = ovl_dentry_lowerdata(dentry); |
74 | if (!lower) | |
d101a125 | 75 | goto bug; |
cef4cbff | 76 | real = lower; |
d101a125 | 77 | |
c4fcfc16 | 78 | /* Handle recursion */ |
fb16043b | 79 | real = d_real(real, inode); |
c4fcfc16 | 80 | |
d101a125 MS |
81 | if (!inode || inode == d_inode(real)) |
82 | return real; | |
d101a125 | 83 | bug: |
cef4cbff MS |
84 | WARN(1, "%s(%pd4, %s:%lu): real dentry (%p/%lu) not found\n", |
85 | __func__, dentry, inode ? inode->i_sb->s_id : "NULL", | |
86 | inode ? inode->i_ino : 0, real, | |
87 | real && d_inode(real) ? d_inode(real)->i_ino : 0); | |
d101a125 MS |
88 | return dentry; |
89 | } | |
90 | ||
3bb7df92 | 91 | static int ovl_revalidate_real(struct dentry *d, unsigned int flags, bool weak) |
7c03b5d4 | 92 | { |
7c03b5d4 MS |
93 | int ret = 1; |
94 | ||
41665644 AG |
95 | if (!d) |
96 | return 1; | |
97 | ||
3bb7df92 MS |
98 | if (weak) { |
99 | if (d->d_flags & DCACHE_OP_WEAK_REVALIDATE) | |
100 | ret = d->d_op->d_weak_revalidate(d, flags); | |
101 | } else if (d->d_flags & DCACHE_OP_REVALIDATE) { | |
102 | ret = d->d_op->d_revalidate(d, flags); | |
103 | if (!ret) { | |
104 | if (!(flags & LOOKUP_RCU)) | |
105 | d_invalidate(d); | |
106 | ret = -ESTALE; | |
7c03b5d4 MS |
107 | } |
108 | } | |
3bb7df92 | 109 | return ret; |
7c03b5d4 MS |
110 | } |
111 | ||
3bb7df92 MS |
112 | static int ovl_dentry_revalidate_common(struct dentry *dentry, |
113 | unsigned int flags, bool weak) | |
7c03b5d4 | 114 | { |
a6ff2bc0 | 115 | struct ovl_entry *oe = OVL_E(dentry); |
5522c9c7 | 116 | struct ovl_path *lowerstack = ovl_lowerstack(oe); |
672e4268 | 117 | struct inode *inode = d_inode_rcu(dentry); |
bccece1e | 118 | struct dentry *upper; |
7c03b5d4 MS |
119 | unsigned int i; |
120 | int ret = 1; | |
121 | ||
672e4268 CZ |
122 | /* Careful in RCU mode */ |
123 | if (!inode) | |
124 | return -ECHILD; | |
125 | ||
126 | upper = ovl_i_dentry_upper(inode); | |
bccece1e MS |
127 | if (upper) |
128 | ret = ovl_revalidate_real(upper, flags, weak); | |
129 | ||
5522c9c7 AG |
130 | for (i = 0; ret > 0 && i < ovl_numlower(oe); i++) |
131 | ret = ovl_revalidate_real(lowerstack[i].dentry, flags, weak); | |
132 | ||
7c03b5d4 MS |
133 | return ret; |
134 | } | |
135 | ||
3bb7df92 MS |
136 | static int ovl_dentry_revalidate(struct dentry *dentry, unsigned int flags) |
137 | { | |
138 | return ovl_dentry_revalidate_common(dentry, flags, false); | |
139 | } | |
140 | ||
141 | static int ovl_dentry_weak_revalidate(struct dentry *dentry, unsigned int flags) | |
142 | { | |
143 | return ovl_dentry_revalidate_common(dentry, flags, true); | |
144 | } | |
145 | ||
e9be9d5e | 146 | static const struct dentry_operations ovl_dentry_operations = { |
d101a125 | 147 | .d_real = ovl_d_real, |
7c03b5d4 MS |
148 | .d_revalidate = ovl_dentry_revalidate, |
149 | .d_weak_revalidate = ovl_dentry_weak_revalidate, | |
150 | }; | |
151 | ||
13cf199d AG |
152 | static struct kmem_cache *ovl_inode_cachep; |
153 | ||
154 | static struct inode *ovl_alloc_inode(struct super_block *sb) | |
155 | { | |
fd60b288 | 156 | struct ovl_inode *oi = alloc_inode_sb(sb, ovl_inode_cachep, GFP_KERNEL); |
13cf199d | 157 | |
b3885bd6 HN |
158 | if (!oi) |
159 | return NULL; | |
160 | ||
04a01ac7 | 161 | oi->cache = NULL; |
cf31c463 | 162 | oi->redirect = NULL; |
04a01ac7 | 163 | oi->version = 0; |
13c72075 | 164 | oi->flags = 0; |
09d8b586 | 165 | oi->__upperdentry = NULL; |
2b21da92 | 166 | oi->lowerdata_redirect = NULL; |
0af950f5 | 167 | oi->oe = NULL; |
a015dafc | 168 | mutex_init(&oi->lock); |
25b7713a | 169 | |
13cf199d AG |
170 | return &oi->vfs_inode; |
171 | } | |
172 | ||
0b269ded | 173 | static void ovl_free_inode(struct inode *inode) |
13cf199d | 174 | { |
0b269ded | 175 | struct ovl_inode *oi = OVL_I(inode); |
13cf199d | 176 | |
0b269ded AV |
177 | kfree(oi->redirect); |
178 | mutex_destroy(&oi->lock); | |
179 | kmem_cache_free(ovl_inode_cachep, oi); | |
13cf199d AG |
180 | } |
181 | ||
182 | static void ovl_destroy_inode(struct inode *inode) | |
183 | { | |
09d8b586 MS |
184 | struct ovl_inode *oi = OVL_I(inode); |
185 | ||
186 | dput(oi->__upperdentry); | |
0af950f5 | 187 | ovl_free_entry(oi->oe); |
2664bd08 VG |
188 | if (S_ISDIR(inode->i_mode)) |
189 | ovl_dir_cache_free(inode); | |
2b21da92 AG |
190 | else |
191 | kfree(oi->lowerdata_redirect); | |
13cf199d AG |
192 | } |
193 | ||
a9075cdb MS |
194 | static void ovl_put_super(struct super_block *sb) |
195 | { | |
f01d0889 | 196 | struct ovl_fs *ofs = OVL_FS(sb); |
a9075cdb | 197 | |
1784fbc2 CB |
198 | if (ofs) |
199 | ovl_free_fs(ofs); | |
a9075cdb MS |
200 | } |
201 | ||
e8d4bfe3 | 202 | /* Sync real dirty inodes in upper filesystem (if it exists) */ |
e593b2bf AG |
203 | static int ovl_sync_fs(struct super_block *sb, int wait) |
204 | { | |
f01d0889 | 205 | struct ovl_fs *ofs = OVL_FS(sb); |
e593b2bf AG |
206 | struct super_block *upper_sb; |
207 | int ret; | |
208 | ||
335d3fc5 SD |
209 | ret = ovl_sync_status(ofs); |
210 | /* | |
211 | * We have to always set the err, because the return value isn't | |
212 | * checked in syncfs, and instead indirectly return an error via | |
213 | * the sb's writeback errseq, which VFS inspects after this call. | |
214 | */ | |
215 | if (ret < 0) { | |
216 | errseq_set(&sb->s_wb_err, -EIO); | |
217 | return -EIO; | |
218 | } | |
219 | ||
220 | if (!ret) | |
221 | return ret; | |
e8d4bfe3 CX |
222 | |
223 | /* | |
32b1924b KK |
224 | * Not called for sync(2) call or an emergency sync (SB_I_SKIP_SYNC). |
225 | * All the super blocks will be iterated, including upper_sb. | |
e8d4bfe3 CX |
226 | * |
227 | * If this is a syncfs(2) call, then we do need to call | |
228 | * sync_filesystem() on upper_sb, but enough if we do it when being | |
229 | * called with wait == 1. | |
230 | */ | |
231 | if (!wait) | |
e593b2bf AG |
232 | return 0; |
233 | ||
08f4c7c8 | 234 | upper_sb = ovl_upper_mnt(ofs)->mnt_sb; |
e8d4bfe3 | 235 | |
e593b2bf | 236 | down_read(&upper_sb->s_umount); |
e8d4bfe3 | 237 | ret = sync_filesystem(upper_sb); |
e593b2bf | 238 | up_read(&upper_sb->s_umount); |
e8d4bfe3 | 239 | |
e593b2bf AG |
240 | return ret; |
241 | } | |
242 | ||
cc259639 AW |
243 | /** |
244 | * ovl_statfs | |
9c5dd803 | 245 | * @dentry: The dentry to query |
cc259639 AW |
246 | * @buf: The struct kstatfs to fill in with stats |
247 | * | |
248 | * Get the filesystem statistics. As writes always target the upper layer | |
4ebc5818 | 249 | * filesystem pass the statfs to the upper filesystem (if it exists) |
cc259639 AW |
250 | */ |
251 | static int ovl_statfs(struct dentry *dentry, struct kstatfs *buf) | |
252 | { | |
b0504bfe AG |
253 | struct super_block *sb = dentry->d_sb; |
254 | struct ovl_fs *ofs = OVL_FS(sb); | |
255 | struct dentry *root_dentry = sb->s_root; | |
cc259639 AW |
256 | struct path path; |
257 | int err; | |
258 | ||
4ebc5818 | 259 | ovl_path_real(root_dentry, &path); |
cc259639 AW |
260 | |
261 | err = vfs_statfs(&path, buf); | |
262 | if (!err) { | |
6b2d5fe4 | 263 | buf->f_namelen = ofs->namelen; |
cc259639 | 264 | buf->f_type = OVERLAYFS_SUPER_MAGIC; |
b0504bfe AG |
265 | if (ovl_has_fsid(ofs)) |
266 | buf->f_fsid = uuid_to_fsid(sb->s_uuid.b); | |
cc259639 AW |
267 | } |
268 | ||
269 | return err; | |
270 | } | |
271 | ||
e9be9d5e | 272 | static const struct super_operations ovl_super_operations = { |
13cf199d | 273 | .alloc_inode = ovl_alloc_inode, |
0b269ded | 274 | .free_inode = ovl_free_inode, |
13cf199d AG |
275 | .destroy_inode = ovl_destroy_inode, |
276 | .drop_inode = generic_delete_inode, | |
e9be9d5e | 277 | .put_super = ovl_put_super, |
e593b2bf | 278 | .sync_fs = ovl_sync_fs, |
cc259639 | 279 | .statfs = ovl_statfs, |
f45827e8 | 280 | .show_options = ovl_show_options, |
e9be9d5e MS |
281 | }; |
282 | ||
e9be9d5e | 283 | #define OVL_WORKDIR_NAME "work" |
02bcd157 | 284 | #define OVL_INDEXDIR_NAME "index" |
e9be9d5e | 285 | |
ad204488 | 286 | static struct dentry *ovl_workdir_create(struct ovl_fs *ofs, |
6b8aa129 | 287 | const char *name, bool persist) |
e9be9d5e | 288 | { |
ad204488 | 289 | struct inode *dir = ofs->workbasedir->d_inode; |
08f4c7c8 | 290 | struct vfsmount *mnt = ovl_upper_mnt(ofs); |
e9be9d5e MS |
291 | struct dentry *work; |
292 | int err; | |
293 | bool retried = false; | |
294 | ||
5955102c | 295 | inode_lock_nested(dir, I_MUTEX_PARENT); |
e9be9d5e | 296 | retry: |
22f289ce | 297 | work = ovl_lookup_upper(ofs, name, ofs->workbasedir, strlen(name)); |
e9be9d5e MS |
298 | |
299 | if (!IS_ERR(work)) { | |
c11b9fdd MS |
300 | struct iattr attr = { |
301 | .ia_valid = ATTR_MODE, | |
32a3d848 | 302 | .ia_mode = S_IFDIR | 0, |
c11b9fdd | 303 | }; |
e9be9d5e MS |
304 | |
305 | if (work->d_inode) { | |
306 | err = -EEXIST; | |
307 | if (retried) | |
308 | goto out_dput; | |
309 | ||
6b8aa129 AG |
310 | if (persist) |
311 | goto out_unlock; | |
312 | ||
e9be9d5e | 313 | retried = true; |
576bb263 | 314 | err = ovl_workdir_cleanup(ofs, dir, mnt, work, 0); |
e9be9d5e | 315 | dput(work); |
235ce9ed AG |
316 | if (err == -EINVAL) { |
317 | work = ERR_PTR(err); | |
318 | goto out_unlock; | |
319 | } | |
e9be9d5e MS |
320 | goto retry; |
321 | } | |
322 | ||
576bb263 | 323 | err = ovl_mkdir_real(ofs, dir, &work, attr.ia_mode); |
1f5573cf MS |
324 | if (err) |
325 | goto out_dput; | |
326 | ||
327 | /* Weird filesystem returning with hashed negative (kernfs)? */ | |
328 | err = -EINVAL; | |
329 | if (d_really_is_negative(work)) | |
330 | goto out_dput; | |
c11b9fdd | 331 | |
cb348edb MS |
332 | /* |
333 | * Try to remove POSIX ACL xattrs from workdir. We are good if: | |
334 | * | |
335 | * a) success (there was a POSIX ACL xattr and was removed) | |
336 | * b) -ENODATA (there was no POSIX ACL xattr) | |
337 | * c) -EOPNOTSUPP (POSIX ACL xattrs are not supported) | |
338 | * | |
339 | * There are various other error values that could effectively | |
340 | * mean that the xattr doesn't exist (e.g. -ERANGE is returned | |
341 | * if the xattr name is too long), but the set of filesystems | |
342 | * allowed as upper are limited to "normal" ones, where checking | |
343 | * for the above two errors is sufficient. | |
344 | */ | |
31acceb9 | 345 | err = ovl_do_remove_acl(ofs, work, XATTR_NAME_POSIX_ACL_DEFAULT); |
e1ff3dd1 | 346 | if (err && err != -ENODATA && err != -EOPNOTSUPP) |
c11b9fdd MS |
347 | goto out_dput; |
348 | ||
31acceb9 | 349 | err = ovl_do_remove_acl(ofs, work, XATTR_NAME_POSIX_ACL_ACCESS); |
e1ff3dd1 | 350 | if (err && err != -ENODATA && err != -EOPNOTSUPP) |
c11b9fdd MS |
351 | goto out_dput; |
352 | ||
353 | /* Clear any inherited mode bits */ | |
354 | inode_lock(work->d_inode); | |
a15506ea | 355 | err = ovl_do_notify_change(ofs, work, &attr); |
c11b9fdd MS |
356 | inode_unlock(work->d_inode); |
357 | if (err) | |
358 | goto out_dput; | |
6b8aa129 AG |
359 | } else { |
360 | err = PTR_ERR(work); | |
361 | goto out_err; | |
e9be9d5e MS |
362 | } |
363 | out_unlock: | |
2068cf7d | 364 | inode_unlock(dir); |
e9be9d5e MS |
365 | return work; |
366 | ||
367 | out_dput: | |
368 | dput(work); | |
6b8aa129 | 369 | out_err: |
1bd0a3ae | 370 | pr_warn("failed to create directory %s/%s (errno: %i); mounting read-only\n", |
ad204488 | 371 | ofs->config.workdir, name, -err); |
6b8aa129 | 372 | work = NULL; |
e9be9d5e MS |
373 | goto out_unlock; |
374 | } | |
375 | ||
2d343087 | 376 | static int ovl_check_namelen(const struct path *path, struct ovl_fs *ofs, |
6b2d5fe4 | 377 | const char *name) |
ab508822 | 378 | { |
ab508822 | 379 | struct kstatfs statfs; |
6b2d5fe4 MS |
380 | int err = vfs_statfs(path, &statfs); |
381 | ||
382 | if (err) | |
1bd0a3ae | 383 | pr_err("statfs failed on '%s'\n", name); |
6b2d5fe4 MS |
384 | else |
385 | ofs->namelen = max(ofs->namelen, statfs.f_namelen); | |
386 | ||
387 | return err; | |
388 | } | |
389 | ||
390 | static int ovl_lower_dir(const char *name, struct path *path, | |
f4288844 | 391 | struct ovl_fs *ofs, int *stack_depth) |
6b2d5fe4 | 392 | { |
e487d889 | 393 | int fh_type; |
6b2d5fe4 | 394 | int err; |
ab508822 | 395 | |
6b2d5fe4 MS |
396 | err = ovl_check_namelen(path, ofs, name); |
397 | if (err) | |
b8e42a65 | 398 | return err; |
6b2d5fe4 | 399 | |
ab508822 MS |
400 | *stack_depth = max(*stack_depth, path->mnt->mnt_sb->s_stack_depth); |
401 | ||
02bcd157 | 402 | /* |
f168f109 AG |
403 | * The inodes index feature and NFS export need to encode and decode |
404 | * file handles, so they require that all layers support them. | |
02bcd157 | 405 | */ |
e487d889 | 406 | fh_type = ovl_can_decode_fh(path->dentry->d_sb); |
f168f109 | 407 | if ((ofs->config.nfs_export || |
e487d889 | 408 | (ofs->config.index && ofs->config.upperdir)) && !fh_type) { |
02bcd157 | 409 | ofs->config.index = false; |
f168f109 | 410 | ofs->config.nfs_export = false; |
1bd0a3ae | 411 | pr_warn("fs on '%s' does not support file handles, falling back to index=off,nfs_export=off.\n", |
f168f109 | 412 | name); |
02bcd157 | 413 | } |
16aac5ad | 414 | ofs->nofh |= !fh_type; |
b0e0f697 AG |
415 | /* |
416 | * Decoding origin file handle is required for persistent st_ino. | |
417 | * Without persistent st_ino, xino=auto falls back to xino=off. | |
418 | */ | |
419 | if (ofs->config.xino == OVL_XINO_AUTO && | |
420 | ofs->config.upperdir && !fh_type) { | |
421 | ofs->config.xino = OVL_XINO_OFF; | |
422 | pr_warn("fs on '%s' does not support file handles, falling back to xino=off.\n", | |
423 | name); | |
424 | } | |
02bcd157 | 425 | |
e487d889 AG |
426 | /* Check if lower fs has 32bit inode numbers */ |
427 | if (fh_type != FILEID_INO32_GEN) | |
0f831ec8 | 428 | ofs->xino_mode = -1; |
e487d889 | 429 | |
ab508822 | 430 | return 0; |
ab508822 MS |
431 | } |
432 | ||
e9be9d5e MS |
433 | /* Workdir should not be subdir of upperdir and vice versa */ |
434 | static bool ovl_workdir_ok(struct dentry *workdir, struct dentry *upperdir) | |
435 | { | |
436 | bool ok = false; | |
437 | ||
438 | if (workdir != upperdir) { | |
439 | ok = (lock_rename(workdir, upperdir) == NULL); | |
440 | unlock_rename(workdir, upperdir); | |
441 | } | |
442 | return ok; | |
443 | } | |
444 | ||
0eb45fc3 AG |
445 | static int ovl_own_xattr_get(const struct xattr_handler *handler, |
446 | struct dentry *dentry, struct inode *inode, | |
447 | const char *name, void *buffer, size_t size) | |
448 | { | |
48fab5d7 | 449 | return -EOPNOTSUPP; |
0eb45fc3 AG |
450 | } |
451 | ||
d837a49b | 452 | static int ovl_own_xattr_set(const struct xattr_handler *handler, |
39f60c1c | 453 | struct mnt_idmap *idmap, |
d837a49b MS |
454 | struct dentry *dentry, struct inode *inode, |
455 | const char *name, const void *value, | |
456 | size_t size, int flags) | |
457 | { | |
48fab5d7 | 458 | return -EOPNOTSUPP; |
d837a49b MS |
459 | } |
460 | ||
0eb45fc3 AG |
461 | static int ovl_other_xattr_get(const struct xattr_handler *handler, |
462 | struct dentry *dentry, struct inode *inode, | |
463 | const char *name, void *buffer, size_t size) | |
464 | { | |
1d88f183 | 465 | return ovl_xattr_get(dentry, inode, name, buffer, size); |
0eb45fc3 AG |
466 | } |
467 | ||
0e585ccc | 468 | static int ovl_other_xattr_set(const struct xattr_handler *handler, |
39f60c1c | 469 | struct mnt_idmap *idmap, |
0e585ccc AG |
470 | struct dentry *dentry, struct inode *inode, |
471 | const char *name, const void *value, | |
472 | size_t size, int flags) | |
473 | { | |
1d88f183 | 474 | return ovl_xattr_set(dentry, inode, name, value, size, flags); |
0e585ccc AG |
475 | } |
476 | ||
2d2f2d73 MS |
477 | static const struct xattr_handler ovl_own_trusted_xattr_handler = { |
478 | .prefix = OVL_XATTR_TRUSTED_PREFIX, | |
479 | .get = ovl_own_xattr_get, | |
480 | .set = ovl_own_xattr_set, | |
481 | }; | |
482 | ||
483 | static const struct xattr_handler ovl_own_user_xattr_handler = { | |
484 | .prefix = OVL_XATTR_USER_PREFIX, | |
0eb45fc3 | 485 | .get = ovl_own_xattr_get, |
d837a49b MS |
486 | .set = ovl_own_xattr_set, |
487 | }; | |
488 | ||
489 | static const struct xattr_handler ovl_other_xattr_handler = { | |
490 | .prefix = "", /* catch all */ | |
0eb45fc3 | 491 | .get = ovl_other_xattr_get, |
d837a49b MS |
492 | .set = ovl_other_xattr_set, |
493 | }; | |
494 | ||
2d2f2d73 | 495 | static const struct xattr_handler *ovl_trusted_xattr_handlers[] = { |
2d2f2d73 MS |
496 | &ovl_own_trusted_xattr_handler, |
497 | &ovl_other_xattr_handler, | |
498 | NULL | |
499 | }; | |
500 | ||
501 | static const struct xattr_handler *ovl_user_xattr_handlers[] = { | |
2d2f2d73 | 502 | &ovl_own_user_xattr_handler, |
d837a49b MS |
503 | &ovl_other_xattr_handler, |
504 | NULL | |
505 | }; | |
506 | ||
146d62e5 AG |
507 | static int ovl_setup_trap(struct super_block *sb, struct dentry *dir, |
508 | struct inode **ptrap, const char *name) | |
509 | { | |
510 | struct inode *trap; | |
511 | int err; | |
512 | ||
513 | trap = ovl_get_trap_inode(sb, dir); | |
1dac6f5b AB |
514 | err = PTR_ERR_OR_ZERO(trap); |
515 | if (err) { | |
146d62e5 | 516 | if (err == -ELOOP) |
1bd0a3ae | 517 | pr_err("conflicting %s path\n", name); |
146d62e5 AG |
518 | return err; |
519 | } | |
520 | ||
521 | *ptrap = trap; | |
522 | return 0; | |
523 | } | |
524 | ||
0be0bfd2 AG |
525 | /* |
526 | * Determine how we treat concurrent use of upperdir/workdir based on the | |
527 | * index feature. This is papering over mount leaks of container runtimes, | |
528 | * for example, an old overlay mount is leaked and now its upperdir is | |
529 | * attempted to be used as a lower layer in a new overlay mount. | |
530 | */ | |
531 | static int ovl_report_in_use(struct ovl_fs *ofs, const char *name) | |
532 | { | |
533 | if (ofs->config.index) { | |
1bd0a3ae | 534 | pr_err("%s is in-use as upperdir/workdir of another mount, mount with '-o index=off' to override exclusive upperdir protection.\n", |
0be0bfd2 AG |
535 | name); |
536 | return -EBUSY; | |
537 | } else { | |
1bd0a3ae | 538 | pr_warn("%s is in-use as upperdir/workdir of another mount, accessing files from both mounts will result in undefined behavior.\n", |
0be0bfd2 AG |
539 | name); |
540 | return 0; | |
541 | } | |
542 | } | |
543 | ||
146d62e5 | 544 | static int ovl_get_upper(struct super_block *sb, struct ovl_fs *ofs, |
b36a5780 CB |
545 | struct ovl_layer *upper_layer, |
546 | const struct path *upperpath) | |
6ee8acf0 | 547 | { |
5064975e | 548 | struct vfsmount *upper_mnt; |
6ee8acf0 MS |
549 | int err; |
550 | ||
e21a6c57 AG |
551 | /* Upperdir path should not be r/o */ |
552 | if (__mnt_is_readonly(upperpath->mnt)) { | |
1bd0a3ae | 553 | pr_err("upper fs is r/o, try multi-lower layers mount\n"); |
6ee8acf0 MS |
554 | err = -EINVAL; |
555 | goto out; | |
556 | } | |
557 | ||
ad204488 | 558 | err = ovl_check_namelen(upperpath, ofs, ofs->config.upperdir); |
6ee8acf0 MS |
559 | if (err) |
560 | goto out; | |
561 | ||
b8e42a65 | 562 | err = ovl_setup_trap(sb, upperpath->dentry, &upper_layer->trap, |
146d62e5 AG |
563 | "upperdir"); |
564 | if (err) | |
565 | goto out; | |
566 | ||
5064975e MS |
567 | upper_mnt = clone_private_mount(upperpath); |
568 | err = PTR_ERR(upper_mnt); | |
569 | if (IS_ERR(upper_mnt)) { | |
1bd0a3ae | 570 | pr_err("failed to clone upperpath\n"); |
5064975e MS |
571 | goto out; |
572 | } | |
573 | ||
574 | /* Don't inherit atime flags */ | |
575 | upper_mnt->mnt_flags &= ~(MNT_NOATIME | MNT_NODIRATIME | MNT_RELATIME); | |
b8e42a65 MS |
576 | upper_layer->mnt = upper_mnt; |
577 | upper_layer->idx = 0; | |
578 | upper_layer->fsid = 0; | |
8c25741a | 579 | |
b36a5780 CB |
580 | err = -ENOMEM; |
581 | upper_layer->name = kstrdup(ofs->config.upperdir, GFP_KERNEL); | |
582 | if (!upper_layer->name) | |
583 | goto out; | |
584 | ||
654255fa JX |
585 | /* |
586 | * Inherit SB_NOSEC flag from upperdir. | |
587 | * | |
588 | * This optimization changes behavior when a security related attribute | |
589 | * (suid/sgid/security.*) is changed on an underlying layer. This is | |
590 | * okay because we don't yet have guarantees in that case, but it will | |
591 | * need careful treatment once we want to honour changes to underlying | |
592 | * filesystems. | |
593 | */ | |
594 | if (upper_mnt->mnt_sb->s_flags & SB_NOSEC) | |
595 | sb->s_flags |= SB_NOSEC; | |
596 | ||
08f4c7c8 | 597 | if (ovl_inuse_trylock(ovl_upper_mnt(ofs)->mnt_root)) { |
8c25741a | 598 | ofs->upperdir_locked = true; |
8c25741a | 599 | } else { |
0be0bfd2 AG |
600 | err = ovl_report_in_use(ofs, "upperdir"); |
601 | if (err) | |
602 | goto out; | |
8c25741a MS |
603 | } |
604 | ||
6ee8acf0 MS |
605 | err = 0; |
606 | out: | |
607 | return err; | |
608 | } | |
609 | ||
cad218ab AG |
610 | /* |
611 | * Returns 1 if RENAME_WHITEOUT is supported, 0 if not supported and | |
612 | * negative values if error is encountered. | |
613 | */ | |
576bb263 | 614 | static int ovl_check_rename_whiteout(struct ovl_fs *ofs) |
cad218ab | 615 | { |
576bb263 | 616 | struct dentry *workdir = ofs->workdir; |
cad218ab AG |
617 | struct inode *dir = d_inode(workdir); |
618 | struct dentry *temp; | |
619 | struct dentry *dest; | |
620 | struct dentry *whiteout; | |
621 | struct name_snapshot name; | |
622 | int err; | |
623 | ||
624 | inode_lock_nested(dir, I_MUTEX_PARENT); | |
625 | ||
576bb263 | 626 | temp = ovl_create_temp(ofs, workdir, OVL_CATTR(S_IFREG | 0)); |
cad218ab AG |
627 | err = PTR_ERR(temp); |
628 | if (IS_ERR(temp)) | |
629 | goto out_unlock; | |
630 | ||
576bb263 | 631 | dest = ovl_lookup_temp(ofs, workdir); |
cad218ab AG |
632 | err = PTR_ERR(dest); |
633 | if (IS_ERR(dest)) { | |
634 | dput(temp); | |
635 | goto out_unlock; | |
636 | } | |
637 | ||
638 | /* Name is inline and stable - using snapshot as a copy helper */ | |
639 | take_dentry_name_snapshot(&name, temp); | |
576bb263 | 640 | err = ovl_do_rename(ofs, dir, temp, dir, dest, RENAME_WHITEOUT); |
cad218ab AG |
641 | if (err) { |
642 | if (err == -EINVAL) | |
643 | err = 0; | |
644 | goto cleanup_temp; | |
645 | } | |
646 | ||
22f289ce | 647 | whiteout = ovl_lookup_upper(ofs, name.name.name, workdir, name.name.len); |
cad218ab AG |
648 | err = PTR_ERR(whiteout); |
649 | if (IS_ERR(whiteout)) | |
650 | goto cleanup_temp; | |
651 | ||
652 | err = ovl_is_whiteout(whiteout); | |
653 | ||
654 | /* Best effort cleanup of whiteout and temp file */ | |
655 | if (err) | |
576bb263 | 656 | ovl_cleanup(ofs, dir, whiteout); |
cad218ab AG |
657 | dput(whiteout); |
658 | ||
659 | cleanup_temp: | |
576bb263 | 660 | ovl_cleanup(ofs, dir, temp); |
cad218ab AG |
661 | release_dentry_name_snapshot(&name); |
662 | dput(temp); | |
663 | dput(dest); | |
664 | ||
665 | out_unlock: | |
666 | inode_unlock(dir); | |
667 | ||
668 | return err; | |
669 | } | |
670 | ||
576bb263 CB |
671 | static struct dentry *ovl_lookup_or_create(struct ovl_fs *ofs, |
672 | struct dentry *parent, | |
c86243b0 VG |
673 | const char *name, umode_t mode) |
674 | { | |
675 | size_t len = strlen(name); | |
676 | struct dentry *child; | |
677 | ||
678 | inode_lock_nested(parent->d_inode, I_MUTEX_PARENT); | |
22f289ce | 679 | child = ovl_lookup_upper(ofs, name, parent, len); |
c86243b0 | 680 | if (!IS_ERR(child) && !child->d_inode) |
576bb263 | 681 | child = ovl_create_real(ofs, parent->d_inode, child, |
c86243b0 VG |
682 | OVL_CATTR(mode)); |
683 | inode_unlock(parent->d_inode); | |
684 | dput(parent); | |
685 | ||
686 | return child; | |
687 | } | |
688 | ||
689 | /* | |
690 | * Creates $workdir/work/incompat/volatile/dirty file if it is not already | |
691 | * present. | |
692 | */ | |
693 | static int ovl_create_volatile_dirty(struct ovl_fs *ofs) | |
694 | { | |
695 | unsigned int ctr; | |
696 | struct dentry *d = dget(ofs->workbasedir); | |
697 | static const char *const volatile_path[] = { | |
698 | OVL_WORKDIR_NAME, "incompat", "volatile", "dirty" | |
699 | }; | |
700 | const char *const *name = volatile_path; | |
701 | ||
702 | for (ctr = ARRAY_SIZE(volatile_path); ctr; ctr--, name++) { | |
576bb263 | 703 | d = ovl_lookup_or_create(ofs, d, *name, ctr > 1 ? S_IFDIR : S_IFREG); |
c86243b0 VG |
704 | if (IS_ERR(d)) |
705 | return PTR_ERR(d); | |
706 | } | |
707 | dput(d); | |
708 | return 0; | |
709 | } | |
710 | ||
146d62e5 | 711 | static int ovl_make_workdir(struct super_block *sb, struct ovl_fs *ofs, |
2d343087 | 712 | const struct path *workpath) |
8ed61dc3 | 713 | { |
08f4c7c8 | 714 | struct vfsmount *mnt = ovl_upper_mnt(ofs); |
2b1a7746 MS |
715 | struct dentry *workdir; |
716 | struct file *tmpfile; | |
d80172c2 AG |
717 | bool rename_whiteout; |
718 | bool d_type; | |
e487d889 | 719 | int fh_type; |
8ed61dc3 MS |
720 | int err; |
721 | ||
2ba9d57e AG |
722 | err = mnt_want_write(mnt); |
723 | if (err) | |
724 | return err; | |
725 | ||
235ce9ed AG |
726 | workdir = ovl_workdir_create(ofs, OVL_WORKDIR_NAME, false); |
727 | err = PTR_ERR(workdir); | |
728 | if (IS_ERR_OR_NULL(workdir)) | |
2ba9d57e | 729 | goto out; |
8ed61dc3 | 730 | |
235ce9ed AG |
731 | ofs->workdir = workdir; |
732 | ||
146d62e5 AG |
733 | err = ovl_setup_trap(sb, ofs->workdir, &ofs->workdir_trap, "workdir"); |
734 | if (err) | |
735 | goto out; | |
736 | ||
8ed61dc3 MS |
737 | /* |
738 | * Upper should support d_type, else whiteouts are visible. Given | |
739 | * workdir and upper are on same fs, we can do iterate_dir() on | |
740 | * workdir. This check requires successful creation of workdir in | |
741 | * previous step. | |
742 | */ | |
743 | err = ovl_check_d_type_supported(workpath); | |
744 | if (err < 0) | |
2ba9d57e | 745 | goto out; |
8ed61dc3 | 746 | |
d80172c2 AG |
747 | d_type = err; |
748 | if (!d_type) | |
1bd0a3ae | 749 | pr_warn("upper fs needs to support d_type.\n"); |
8ed61dc3 MS |
750 | |
751 | /* Check if upper/work fs supports O_TMPFILE */ | |
2b1a7746 MS |
752 | tmpfile = ovl_do_tmpfile(ofs, ofs->workdir, S_IFREG | 0); |
753 | ofs->tmpfile = !IS_ERR(tmpfile); | |
ad204488 | 754 | if (ofs->tmpfile) |
2b1a7746 | 755 | fput(tmpfile); |
8ed61dc3 | 756 | else |
1bd0a3ae | 757 | pr_warn("upper fs does not support tmpfile.\n"); |
8ed61dc3 | 758 | |
cad218ab AG |
759 | |
760 | /* Check if upper/work fs supports RENAME_WHITEOUT */ | |
576bb263 | 761 | err = ovl_check_rename_whiteout(ofs); |
cad218ab AG |
762 | if (err < 0) |
763 | goto out; | |
764 | ||
d80172c2 AG |
765 | rename_whiteout = err; |
766 | if (!rename_whiteout) | |
cad218ab AG |
767 | pr_warn("upper fs does not support RENAME_WHITEOUT.\n"); |
768 | ||
8ed61dc3 | 769 | /* |
2d2f2d73 | 770 | * Check if upper/work fs supports (trusted|user).overlay.* xattr |
8ed61dc3 | 771 | */ |
c914c0e2 | 772 | err = ovl_setxattr(ofs, ofs->workdir, OVL_XATTR_OPAQUE, "0", 1); |
8ed61dc3 | 773 | if (err) { |
b10b85fe | 774 | pr_warn("failed to set xattr on upper\n"); |
ad204488 | 775 | ofs->noxattr = true; |
af5f2396 AG |
776 | if (ovl_redirect_follow(ofs)) { |
777 | ofs->config.redirect_mode = OVL_REDIRECT_NOFOLLOW; | |
778 | pr_warn("...falling back to redirect_dir=nofollow.\n"); | |
779 | } | |
780 | if (ofs->config.metacopy) { | |
b0e0f697 | 781 | ofs->config.metacopy = false; |
af5f2396 AG |
782 | pr_warn("...falling back to metacopy=off.\n"); |
783 | } | |
784 | if (ofs->config.index) { | |
785 | ofs->config.index = false; | |
786 | pr_warn("...falling back to index=off.\n"); | |
b0e0f697 | 787 | } |
d9544c1b AG |
788 | if (ovl_has_fsid(ofs)) { |
789 | ofs->config.uuid = OVL_UUID_NULL; | |
790 | pr_warn("...falling back to uuid=null.\n"); | |
791 | } | |
b0e0f697 AG |
792 | /* |
793 | * xattr support is required for persistent st_ino. | |
794 | * Without persistent st_ino, xino=auto falls back to xino=off. | |
795 | */ | |
796 | if (ofs->config.xino == OVL_XINO_AUTO) { | |
797 | ofs->config.xino = OVL_XINO_OFF; | |
b10b85fe | 798 | pr_warn("...falling back to xino=off.\n"); |
b0e0f697 | 799 | } |
b10b85fe MS |
800 | if (err == -EPERM && !ofs->config.userxattr) |
801 | pr_info("try mounting with 'userxattr' option\n"); | |
2ba9d57e | 802 | err = 0; |
8ed61dc3 | 803 | } else { |
c914c0e2 | 804 | ovl_removexattr(ofs, ofs->workdir, OVL_XATTR_OPAQUE); |
8ed61dc3 MS |
805 | } |
806 | ||
d80172c2 AG |
807 | /* |
808 | * We allowed sub-optimal upper fs configuration and don't want to break | |
809 | * users over kernel upgrade, but we never allowed remote upper fs, so | |
810 | * we can enforce strict requirements for remote upper fs. | |
811 | */ | |
812 | if (ovl_dentry_remote(ofs->workdir) && | |
813 | (!d_type || !rename_whiteout || ofs->noxattr)) { | |
814 | pr_err("upper fs missing required features.\n"); | |
815 | err = -EINVAL; | |
816 | goto out; | |
817 | } | |
818 | ||
c86243b0 VG |
819 | /* |
820 | * For volatile mount, create a incompat/volatile/dirty file to keep | |
821 | * track of it. | |
822 | */ | |
823 | if (ofs->config.ovl_volatile) { | |
824 | err = ovl_create_volatile_dirty(ofs); | |
825 | if (err < 0) { | |
826 | pr_err("Failed to create volatile/dirty file.\n"); | |
827 | goto out; | |
828 | } | |
829 | } | |
830 | ||
8ed61dc3 | 831 | /* Check if upper/work fs supports file handles */ |
e487d889 AG |
832 | fh_type = ovl_can_decode_fh(ofs->workdir->d_sb); |
833 | if (ofs->config.index && !fh_type) { | |
ad204488 | 834 | ofs->config.index = false; |
1bd0a3ae | 835 | pr_warn("upper fs does not support file handles, falling back to index=off.\n"); |
8ed61dc3 | 836 | } |
16aac5ad | 837 | ofs->nofh |= !fh_type; |
8ed61dc3 | 838 | |
e487d889 AG |
839 | /* Check if upper fs has 32bit inode numbers */ |
840 | if (fh_type != FILEID_INO32_GEN) | |
0f831ec8 | 841 | ofs->xino_mode = -1; |
e487d889 | 842 | |
f168f109 AG |
843 | /* NFS export of r/w mount depends on index */ |
844 | if (ofs->config.nfs_export && !ofs->config.index) { | |
1bd0a3ae | 845 | pr_warn("NFS export requires \"index=on\", falling back to nfs_export=off.\n"); |
f168f109 AG |
846 | ofs->config.nfs_export = false; |
847 | } | |
2ba9d57e AG |
848 | out: |
849 | mnt_drop_write(mnt); | |
850 | return err; | |
8ed61dc3 MS |
851 | } |
852 | ||
146d62e5 | 853 | static int ovl_get_workdir(struct super_block *sb, struct ovl_fs *ofs, |
b36a5780 CB |
854 | const struct path *upperpath, |
855 | const struct path *workpath) | |
520d7c86 MS |
856 | { |
857 | int err; | |
520d7c86 MS |
858 | |
859 | err = -EINVAL; | |
b36a5780 | 860 | if (upperpath->mnt != workpath->mnt) { |
1bd0a3ae | 861 | pr_err("workdir and upperdir must reside under the same mount\n"); |
b36a5780 | 862 | return err; |
520d7c86 | 863 | } |
b36a5780 | 864 | if (!ovl_workdir_ok(workpath->dentry, upperpath->dentry)) { |
1bd0a3ae | 865 | pr_err("workdir and upperdir must be separate subtrees\n"); |
b36a5780 | 866 | return err; |
520d7c86 MS |
867 | } |
868 | ||
b36a5780 | 869 | ofs->workbasedir = dget(workpath->dentry); |
8c25741a | 870 | |
8c25741a | 871 | if (ovl_inuse_trylock(ofs->workbasedir)) { |
ad204488 | 872 | ofs->workdir_locked = true; |
520d7c86 | 873 | } else { |
0be0bfd2 AG |
874 | err = ovl_report_in_use(ofs, "workdir"); |
875 | if (err) | |
b36a5780 | 876 | return err; |
520d7c86 MS |
877 | } |
878 | ||
0be0bfd2 AG |
879 | err = ovl_setup_trap(sb, ofs->workbasedir, &ofs->workbasedir_trap, |
880 | "workdir"); | |
881 | if (err) | |
b36a5780 | 882 | return err; |
bca44b52 | 883 | |
b36a5780 | 884 | return ovl_make_workdir(sb, ofs, workpath); |
520d7c86 MS |
885 | } |
886 | ||
146d62e5 | 887 | static int ovl_get_indexdir(struct super_block *sb, struct ovl_fs *ofs, |
2d343087 | 888 | struct ovl_entry *oe, const struct path *upperpath) |
f7e3a7d9 | 889 | { |
08f4c7c8 | 890 | struct vfsmount *mnt = ovl_upper_mnt(ofs); |
235ce9ed | 891 | struct dentry *indexdir; |
f7e3a7d9 MS |
892 | int err; |
893 | ||
2ba9d57e AG |
894 | err = mnt_want_write(mnt); |
895 | if (err) | |
896 | return err; | |
897 | ||
f7e3a7d9 | 898 | /* Verify lower root is upper root origin */ |
610afc0b | 899 | err = ovl_verify_origin(ofs, upperpath->dentry, |
5522c9c7 | 900 | ovl_lowerstack(oe)->dentry, true); |
f7e3a7d9 | 901 | if (err) { |
1bd0a3ae | 902 | pr_err("failed to verify upper root origin\n"); |
f7e3a7d9 MS |
903 | goto out; |
904 | } | |
905 | ||
470c1563 AG |
906 | /* index dir will act also as workdir */ |
907 | iput(ofs->workdir_trap); | |
908 | ofs->workdir_trap = NULL; | |
909 | dput(ofs->workdir); | |
910 | ofs->workdir = NULL; | |
235ce9ed AG |
911 | indexdir = ovl_workdir_create(ofs, OVL_INDEXDIR_NAME, true); |
912 | if (IS_ERR(indexdir)) { | |
913 | err = PTR_ERR(indexdir); | |
914 | } else if (indexdir) { | |
915 | ofs->indexdir = indexdir; | |
916 | ofs->workdir = dget(indexdir); | |
20396365 | 917 | |
146d62e5 AG |
918 | err = ovl_setup_trap(sb, ofs->indexdir, &ofs->indexdir_trap, |
919 | "indexdir"); | |
920 | if (err) | |
921 | goto out; | |
922 | ||
ad1d615c AG |
923 | /* |
924 | * Verify upper root is exclusively associated with index dir. | |
2d2f2d73 | 925 | * Older kernels stored upper fh in ".overlay.origin" |
ad1d615c AG |
926 | * xattr. If that xattr exists, verify that it is a match to |
927 | * upper dir file handle. In any case, verify or set xattr | |
2d2f2d73 | 928 | * ".overlay.upper" to indicate that index may have |
ad1d615c AG |
929 | * directory entries. |
930 | */ | |
610afc0b MS |
931 | if (ovl_check_origin_xattr(ofs, ofs->indexdir)) { |
932 | err = ovl_verify_set_fh(ofs, ofs->indexdir, | |
933 | OVL_XATTR_ORIGIN, | |
ad1d615c AG |
934 | upperpath->dentry, true, false); |
935 | if (err) | |
1bd0a3ae | 936 | pr_err("failed to verify index dir 'origin' xattr\n"); |
ad1d615c | 937 | } |
610afc0b MS |
938 | err = ovl_verify_upper(ofs, ofs->indexdir, upperpath->dentry, |
939 | true); | |
f7e3a7d9 | 940 | if (err) |
1bd0a3ae | 941 | pr_err("failed to verify index dir 'upper' xattr\n"); |
f7e3a7d9 MS |
942 | |
943 | /* Cleanup bad/stale/orphan index entries */ | |
944 | if (!err) | |
1eff1a1d | 945 | err = ovl_indexdir_cleanup(ofs); |
f7e3a7d9 | 946 | } |
ad204488 | 947 | if (err || !ofs->indexdir) |
1bd0a3ae | 948 | pr_warn("try deleting index dir or mounting with '-o index=off' to disable inodes index.\n"); |
f7e3a7d9 MS |
949 | |
950 | out: | |
2ba9d57e | 951 | mnt_drop_write(mnt); |
f7e3a7d9 MS |
952 | return err; |
953 | } | |
954 | ||
9df085f3 AG |
955 | static bool ovl_lower_uuid_ok(struct ovl_fs *ofs, const uuid_t *uuid) |
956 | { | |
957 | unsigned int i; | |
958 | ||
08f4c7c8 | 959 | if (!ofs->config.nfs_export && !ovl_upper_mnt(ofs)) |
9df085f3 AG |
960 | return true; |
961 | ||
a888db31 AG |
962 | /* |
963 | * We allow using single lower with null uuid for index and nfs_export | |
964 | * for example to support those features with single lower squashfs. | |
965 | * To avoid regressions in setups of overlay with re-formatted lower | |
966 | * squashfs, do not allow decoding origin with lower null uuid unless | |
967 | * user opted-in to one of the new features that require following the | |
968 | * lower inode of non-dir upper. | |
969 | */ | |
ca45275c | 970 | if (ovl_allow_offline_changes(ofs) && uuid_is_null(uuid)) |
a888db31 AG |
971 | return false; |
972 | ||
1b81dddd | 973 | for (i = 0; i < ofs->numfs; i++) { |
9df085f3 AG |
974 | /* |
975 | * We use uuid to associate an overlay lower file handle with a | |
976 | * lower layer, so we can accept lower fs with null uuid as long | |
977 | * as all lower layers with null uuid are on the same fs. | |
7e63c87f AG |
978 | * if we detect multiple lower fs with the same uuid, we |
979 | * disable lower file handle decoding on all of them. | |
9df085f3 | 980 | */ |
1b81dddd AG |
981 | if (ofs->fs[i].is_lower && |
982 | uuid_equal(&ofs->fs[i].sb->s_uuid, uuid)) { | |
07f1e596 | 983 | ofs->fs[i].bad_uuid = true; |
9df085f3 | 984 | return false; |
7e63c87f | 985 | } |
9df085f3 AG |
986 | } |
987 | return true; | |
988 | } | |
989 | ||
5148626b | 990 | /* Get a unique fsid for the layer */ |
9df085f3 | 991 | static int ovl_get_fsid(struct ovl_fs *ofs, const struct path *path) |
5148626b | 992 | { |
9df085f3 | 993 | struct super_block *sb = path->mnt->mnt_sb; |
5148626b AG |
994 | unsigned int i; |
995 | dev_t dev; | |
996 | int err; | |
7e63c87f | 997 | bool bad_uuid = false; |
b0e0f697 | 998 | bool warn = false; |
5148626b | 999 | |
07f1e596 AG |
1000 | for (i = 0; i < ofs->numfs; i++) { |
1001 | if (ofs->fs[i].sb == sb) | |
1002 | return i; | |
5148626b AG |
1003 | } |
1004 | ||
9df085f3 | 1005 | if (!ovl_lower_uuid_ok(ofs, &sb->s_uuid)) { |
7e63c87f | 1006 | bad_uuid = true; |
b0e0f697 AG |
1007 | if (ofs->config.xino == OVL_XINO_AUTO) { |
1008 | ofs->config.xino = OVL_XINO_OFF; | |
1009 | warn = true; | |
1010 | } | |
7e63c87f AG |
1011 | if (ofs->config.index || ofs->config.nfs_export) { |
1012 | ofs->config.index = false; | |
1013 | ofs->config.nfs_export = false; | |
b0e0f697 AG |
1014 | warn = true; |
1015 | } | |
1016 | if (warn) { | |
1017 | pr_warn("%s uuid detected in lower fs '%pd2', falling back to xino=%s,index=off,nfs_export=off.\n", | |
7e63c87f AG |
1018 | uuid_is_null(&sb->s_uuid) ? "null" : |
1019 | "conflicting", | |
dcb399de | 1020 | path->dentry, ovl_xino_mode(&ofs->config)); |
7e63c87f | 1021 | } |
9df085f3 AG |
1022 | } |
1023 | ||
5148626b AG |
1024 | err = get_anon_bdev(&dev); |
1025 | if (err) { | |
1bd0a3ae | 1026 | pr_err("failed to get anonymous bdev for lowerpath\n"); |
5148626b AG |
1027 | return err; |
1028 | } | |
1029 | ||
07f1e596 AG |
1030 | ofs->fs[ofs->numfs].sb = sb; |
1031 | ofs->fs[ofs->numfs].pseudo_dev = dev; | |
1032 | ofs->fs[ofs->numfs].bad_uuid = bad_uuid; | |
5148626b | 1033 | |
07f1e596 | 1034 | return ofs->numfs++; |
5148626b AG |
1035 | } |
1036 | ||
37ebf056 AG |
1037 | /* |
1038 | * The fsid after the last lower fsid is used for the data layers. | |
1039 | * It is a "null fs" with a null sb, null uuid, and no pseudo dev. | |
1040 | */ | |
1041 | static int ovl_get_data_fsid(struct ovl_fs *ofs) | |
1042 | { | |
1043 | return ofs->numfs; | |
1044 | } | |
1045 | ||
1046 | ||
94375f9d | 1047 | static int ovl_get_layers(struct super_block *sb, struct ovl_fs *ofs, |
b36a5780 | 1048 | struct ovl_fs_context *ctx, struct ovl_layer *layers) |
520d7c86 MS |
1049 | { |
1050 | int err; | |
1051 | unsigned int i; | |
b36a5780 | 1052 | size_t nr_merged_lower; |
520d7c86 | 1053 | |
b36a5780 | 1054 | ofs->fs = kcalloc(ctx->nr + 2, sizeof(struct ovl_sb), GFP_KERNEL); |
07f1e596 | 1055 | if (ofs->fs == NULL) |
9e88f905 | 1056 | return -ENOMEM; |
5148626b | 1057 | |
37ebf056 AG |
1058 | /* |
1059 | * idx/fsid 0 are reserved for upper fs even with lower only overlay | |
1060 | * and the last fsid is reserved for "null fs" of the data layers. | |
1061 | */ | |
07f1e596 AG |
1062 | ofs->numfs++; |
1063 | ||
07f1e596 | 1064 | /* |
b7bf9908 AG |
1065 | * All lower layers that share the same fs as upper layer, use the same |
1066 | * pseudo_dev as upper layer. Allocate fs[0].pseudo_dev even for lower | |
1067 | * only overlay to simplify ovl_fs_free(). | |
1b81dddd | 1068 | * is_lower will be set if upper fs is shared with a lower layer. |
07f1e596 | 1069 | */ |
b7bf9908 AG |
1070 | err = get_anon_bdev(&ofs->fs[0].pseudo_dev); |
1071 | if (err) { | |
1072 | pr_err("failed to get anonymous bdev for upper fs\n"); | |
9e88f905 | 1073 | return err; |
b7bf9908 AG |
1074 | } |
1075 | ||
08f4c7c8 MS |
1076 | if (ovl_upper_mnt(ofs)) { |
1077 | ofs->fs[0].sb = ovl_upper_mnt(ofs)->mnt_sb; | |
1b81dddd | 1078 | ofs->fs[0].is_lower = false; |
07f1e596 AG |
1079 | } |
1080 | ||
b36a5780 CB |
1081 | nr_merged_lower = ctx->nr - ctx->nr_data; |
1082 | for (i = 0; i < ctx->nr; i++) { | |
1083 | struct ovl_fs_context_layer *l = &ctx->lower[i]; | |
520d7c86 | 1084 | struct vfsmount *mnt; |
146d62e5 | 1085 | struct inode *trap; |
5148626b | 1086 | int fsid; |
520d7c86 | 1087 | |
b36a5780 CB |
1088 | if (i < nr_merged_lower) |
1089 | fsid = ovl_get_fsid(ofs, &l->path); | |
37ebf056 AG |
1090 | else |
1091 | fsid = ovl_get_data_fsid(ofs); | |
9e88f905 AG |
1092 | if (fsid < 0) |
1093 | return fsid; | |
520d7c86 | 1094 | |
24f14009 | 1095 | /* |
1096 | * Check if lower root conflicts with this overlay layers before | |
1097 | * checking if it is in-use as upperdir/workdir of "another" | |
1098 | * mount, because we do not bother to check in ovl_is_inuse() if | |
1099 | * the upperdir/workdir is in fact in-use by our | |
1100 | * upperdir/workdir. | |
1101 | */ | |
b36a5780 | 1102 | err = ovl_setup_trap(sb, l->path.dentry, &trap, "lowerdir"); |
146d62e5 | 1103 | if (err) |
9e88f905 | 1104 | return err; |
146d62e5 | 1105 | |
b36a5780 | 1106 | if (ovl_is_inuse(l->path.dentry)) { |
0be0bfd2 | 1107 | err = ovl_report_in_use(ofs, "lowerdir"); |
24f14009 | 1108 | if (err) { |
1109 | iput(trap); | |
9e88f905 | 1110 | return err; |
24f14009 | 1111 | } |
0be0bfd2 AG |
1112 | } |
1113 | ||
b36a5780 | 1114 | mnt = clone_private_mount(&l->path); |
520d7c86 MS |
1115 | err = PTR_ERR(mnt); |
1116 | if (IS_ERR(mnt)) { | |
1bd0a3ae | 1117 | pr_err("failed to clone lowerpath\n"); |
146d62e5 | 1118 | iput(trap); |
9e88f905 | 1119 | return err; |
520d7c86 | 1120 | } |
5148626b | 1121 | |
520d7c86 MS |
1122 | /* |
1123 | * Make lower layers R/O. That way fchmod/fchown on lower file | |
1124 | * will fail instead of modifying lower fs. | |
1125 | */ | |
1126 | mnt->mnt_flags |= MNT_READONLY | MNT_NOATIME; | |
1127 | ||
13464165 MS |
1128 | layers[ofs->numlayer].trap = trap; |
1129 | layers[ofs->numlayer].mnt = mnt; | |
1130 | layers[ofs->numlayer].idx = ofs->numlayer; | |
1131 | layers[ofs->numlayer].fsid = fsid; | |
1132 | layers[ofs->numlayer].fs = &ofs->fs[fsid]; | |
b36a5780 CB |
1133 | layers[ofs->numlayer].name = l->name; |
1134 | l->name = NULL; | |
94375f9d | 1135 | ofs->numlayer++; |
1b81dddd | 1136 | ofs->fs[fsid].is_lower = true; |
520d7c86 | 1137 | } |
e487d889 | 1138 | |
795939a9 AG |
1139 | /* |
1140 | * When all layers on same fs, overlay can use real inode numbers. | |
926e94d7 AG |
1141 | * With mount option "xino=<on|auto>", mounter declares that there are |
1142 | * enough free high bits in underlying fs to hold the unique fsid. | |
795939a9 AG |
1143 | * If overlayfs does encounter underlying inodes using the high xino |
1144 | * bits reserved for fsid, it emits a warning and uses the original | |
dfe51d47 AG |
1145 | * inode number or a non persistent inode number allocated from a |
1146 | * dedicated range. | |
795939a9 | 1147 | */ |
08f4c7c8 | 1148 | if (ofs->numfs - !ovl_upper_mnt(ofs) == 1) { |
0f831ec8 AG |
1149 | if (ofs->config.xino == OVL_XINO_ON) |
1150 | pr_info("\"xino=on\" is useless with all layers on same fs, ignore.\n"); | |
1151 | ofs->xino_mode = 0; | |
53afcd31 AG |
1152 | } else if (ofs->config.xino == OVL_XINO_OFF) { |
1153 | ofs->xino_mode = -1; | |
926e94d7 | 1154 | } else if (ofs->xino_mode < 0) { |
795939a9 | 1155 | /* |
07f1e596 | 1156 | * This is a roundup of number of bits needed for encoding |
dfe51d47 AG |
1157 | * fsid, where fsid 0 is reserved for upper fs (even with |
1158 | * lower only overlay) +1 extra bit is reserved for the non | |
1159 | * persistent inode number range that is used for resolving | |
1160 | * xino lower bits overflow. | |
795939a9 | 1161 | */ |
dfe51d47 AG |
1162 | BUILD_BUG_ON(ilog2(OVL_MAX_STACK) > 30); |
1163 | ofs->xino_mode = ilog2(ofs->numfs - 1) + 2; | |
795939a9 AG |
1164 | } |
1165 | ||
0f831ec8 | 1166 | if (ofs->xino_mode > 0) { |
1bd0a3ae | 1167 | pr_info("\"xino\" feature enabled using %d upper inode bits.\n", |
0f831ec8 | 1168 | ofs->xino_mode); |
795939a9 | 1169 | } |
e487d889 | 1170 | |
9e88f905 | 1171 | return 0; |
520d7c86 MS |
1172 | } |
1173 | ||
4155c10a | 1174 | static struct ovl_entry *ovl_get_lowerstack(struct super_block *sb, |
b36a5780 CB |
1175 | struct ovl_fs_context *ctx, |
1176 | struct ovl_fs *ofs, | |
1177 | struct ovl_layer *layers) | |
53dbb0b4 MS |
1178 | { |
1179 | int err; | |
b8e42a65 | 1180 | unsigned int i; |
b36a5780 | 1181 | size_t nr_merged_lower; |
4155c10a | 1182 | struct ovl_entry *oe; |
b36a5780 CB |
1183 | struct ovl_path *lowerstack; |
1184 | ||
1185 | struct ovl_fs_context_layer *l; | |
53dbb0b4 | 1186 | |
b36a5780 | 1187 | if (!ofs->config.upperdir && ctx->nr == 1) { |
1bd0a3ae | 1188 | pr_err("at least 2 lowerdir are needed while upperdir nonexistent\n"); |
b8e42a65 | 1189 | return ERR_PTR(-EINVAL); |
53dbb0b4 MS |
1190 | } |
1191 | ||
b36a5780 CB |
1192 | err = -EINVAL; |
1193 | for (i = 0; i < ctx->nr; i++) { | |
1194 | l = &ctx->lower[i]; | |
53dbb0b4 | 1195 | |
b36a5780 | 1196 | err = ovl_lower_dir(l->name, &l->path, ofs, &sb->s_stack_depth); |
53dbb0b4 | 1197 | if (err) |
b36a5780 | 1198 | return ERR_PTR(err); |
53dbb0b4 MS |
1199 | } |
1200 | ||
1201 | err = -EINVAL; | |
1202 | sb->s_stack_depth++; | |
1203 | if (sb->s_stack_depth > FILESYSTEM_MAX_STACK_DEPTH) { | |
1bd0a3ae | 1204 | pr_err("maximum fs stacking depth exceeded\n"); |
b36a5780 | 1205 | return ERR_PTR(err); |
53dbb0b4 MS |
1206 | } |
1207 | ||
b36a5780 | 1208 | err = ovl_get_layers(sb, ofs, ctx, layers); |
4155c10a | 1209 | if (err) |
b36a5780 | 1210 | return ERR_PTR(err); |
4155c10a MS |
1211 | |
1212 | err = -ENOMEM; | |
37ebf056 | 1213 | /* Data-only layers are not merged in root directory */ |
b36a5780 CB |
1214 | nr_merged_lower = ctx->nr - ctx->nr_data; |
1215 | oe = ovl_alloc_entry(nr_merged_lower); | |
4155c10a | 1216 | if (!oe) |
b36a5780 | 1217 | return ERR_PTR(err); |
4155c10a | 1218 | |
5522c9c7 | 1219 | lowerstack = ovl_lowerstack(oe); |
b36a5780 CB |
1220 | for (i = 0; i < nr_merged_lower; i++) { |
1221 | l = &ctx->lower[i]; | |
1222 | lowerstack[i].dentry = dget(l->path.dentry); | |
1223 | lowerstack[i].layer = &ofs->layers[i + 1]; | |
4155c10a | 1224 | } |
b36a5780 | 1225 | ofs->numdatalayer = ctx->nr_data; |
4155c10a MS |
1226 | |
1227 | return oe; | |
53dbb0b4 MS |
1228 | } |
1229 | ||
146d62e5 AG |
1230 | /* |
1231 | * Check if this layer root is a descendant of: | |
1232 | * - another layer of this overlayfs instance | |
1233 | * - upper/work dir of any overlayfs instance | |
146d62e5 | 1234 | */ |
0be0bfd2 | 1235 | static int ovl_check_layer(struct super_block *sb, struct ovl_fs *ofs, |
708fa015 MS |
1236 | struct dentry *dentry, const char *name, |
1237 | bool is_lower) | |
146d62e5 | 1238 | { |
9179c21d | 1239 | struct dentry *next = dentry, *parent; |
146d62e5 AG |
1240 | int err = 0; |
1241 | ||
9179c21d | 1242 | if (!dentry) |
146d62e5 AG |
1243 | return 0; |
1244 | ||
9179c21d MS |
1245 | parent = dget_parent(next); |
1246 | ||
1247 | /* Walk back ancestors to root (inclusive) looking for traps */ | |
1248 | while (!err && parent != next) { | |
708fa015 | 1249 | if (is_lower && ovl_lookup_trap_inode(sb, parent)) { |
146d62e5 | 1250 | err = -ELOOP; |
1bd0a3ae | 1251 | pr_err("overlapping %s path\n", name); |
0be0bfd2 AG |
1252 | } else if (ovl_is_inuse(parent)) { |
1253 | err = ovl_report_in_use(ofs, name); | |
146d62e5 | 1254 | } |
146d62e5 | 1255 | next = parent; |
9179c21d MS |
1256 | parent = dget_parent(next); |
1257 | dput(next); | |
146d62e5 AG |
1258 | } |
1259 | ||
9179c21d | 1260 | dput(parent); |
146d62e5 AG |
1261 | |
1262 | return err; | |
1263 | } | |
1264 | ||
1265 | /* | |
1266 | * Check if any of the layers or work dirs overlap. | |
1267 | */ | |
1268 | static int ovl_check_overlapping_layers(struct super_block *sb, | |
1269 | struct ovl_fs *ofs) | |
1270 | { | |
1271 | int i, err; | |
1272 | ||
08f4c7c8 MS |
1273 | if (ovl_upper_mnt(ofs)) { |
1274 | err = ovl_check_layer(sb, ofs, ovl_upper_mnt(ofs)->mnt_root, | |
708fa015 | 1275 | "upperdir", false); |
146d62e5 AG |
1276 | if (err) |
1277 | return err; | |
1278 | ||
1279 | /* | |
1280 | * Checking workbasedir avoids hitting ovl_is_inuse(parent) of | |
1281 | * this instance and covers overlapping work and index dirs, | |
1282 | * unless work or index dir have been moved since created inside | |
1283 | * workbasedir. In that case, we already have their traps in | |
1284 | * inode cache and we will catch that case on lookup. | |
1285 | */ | |
708fa015 MS |
1286 | err = ovl_check_layer(sb, ofs, ofs->workbasedir, "workdir", |
1287 | false); | |
146d62e5 AG |
1288 | if (err) |
1289 | return err; | |
1290 | } | |
1291 | ||
94375f9d | 1292 | for (i = 1; i < ofs->numlayer; i++) { |
0be0bfd2 | 1293 | err = ovl_check_layer(sb, ofs, |
94375f9d | 1294 | ofs->layers[i].mnt->mnt_root, |
708fa015 | 1295 | "lowerdir", true); |
146d62e5 AG |
1296 | if (err) |
1297 | return err; | |
1298 | } | |
1299 | ||
1300 | return 0; | |
1301 | } | |
1302 | ||
2effc5c2 AG |
1303 | static struct dentry *ovl_get_root(struct super_block *sb, |
1304 | struct dentry *upperdentry, | |
1305 | struct ovl_entry *oe) | |
1306 | { | |
1307 | struct dentry *root; | |
5522c9c7 | 1308 | struct ovl_path *lowerpath = ovl_lowerstack(oe); |
62c832ed AG |
1309 | unsigned long ino = d_inode(lowerpath->dentry)->i_ino; |
1310 | int fsid = lowerpath->layer->fsid; | |
1311 | struct ovl_inode_params oip = { | |
1312 | .upperdentry = upperdentry, | |
0af950f5 | 1313 | .oe = oe, |
62c832ed | 1314 | }; |
2effc5c2 AG |
1315 | |
1316 | root = d_make_root(ovl_new_inode(sb, S_IFDIR, 0)); | |
1317 | if (!root) | |
1318 | return NULL; | |
1319 | ||
2effc5c2 | 1320 | if (upperdentry) { |
62c832ed AG |
1321 | /* Root inode uses upper st_ino/i_ino */ |
1322 | ino = d_inode(upperdentry)->i_ino; | |
1323 | fsid = 0; | |
2effc5c2 | 1324 | ovl_dentry_set_upper_alias(root); |
610afc0b | 1325 | if (ovl_is_impuredir(sb, upperdentry)) |
2effc5c2 AG |
1326 | ovl_set_flag(OVL_IMPURE, d_inode(root)); |
1327 | } | |
1328 | ||
1329 | /* Root is always merge -> can have whiteouts */ | |
1330 | ovl_set_flag(OVL_WHITEOUTS, d_inode(root)); | |
1331 | ovl_dentry_set_flag(OVL_E_CONNECTED, root); | |
1332 | ovl_set_upperdata(d_inode(root)); | |
62c832ed | 1333 | ovl_inode_init(d_inode(root), &oip, ino, fsid); |
0af950f5 | 1334 | ovl_dentry_init_flags(root, upperdentry, oe, DCACHE_OP_WEAK_REVALIDATE); |
367d002d AG |
1335 | /* root keeps a reference of upperdentry */ |
1336 | dget(upperdentry); | |
2effc5c2 AG |
1337 | |
1338 | return root; | |
1339 | } | |
1340 | ||
7fb7998b | 1341 | int ovl_fill_super(struct super_block *sb, struct fs_context *fc) |
e9be9d5e | 1342 | { |
1784fbc2 CB |
1343 | struct ovl_fs *ofs = sb->s_fs_info; |
1344 | struct ovl_fs_context *ctx = fc->fs_private; | |
e9be9d5e | 1345 | struct dentry *root_dentry; |
4155c10a | 1346 | struct ovl_entry *oe; |
b8e42a65 | 1347 | struct ovl_layer *layers; |
51f8f3c4 | 1348 | struct cred *cred; |
e9be9d5e MS |
1349 | int err; |
1350 | ||
9efb069d | 1351 | err = -EIO; |
1784fbc2 CB |
1352 | if (WARN_ON(fc->user_ns != current_user_ns())) |
1353 | goto out_err; | |
9efb069d | 1354 | |
f4288844 MS |
1355 | sb->s_d_op = &ovl_dentry_operations; |
1356 | ||
d7b49b10 | 1357 | err = -ENOMEM; |
ad204488 | 1358 | ofs->creator_cred = cred = prepare_creds(); |
c6fe6254 MS |
1359 | if (!cred) |
1360 | goto out_err; | |
1361 | ||
1784fbc2 | 1362 | err = ovl_fs_params_verify(ctx, &ofs->config); |
f45827e8 | 1363 | if (err) |
a9075cdb | 1364 | goto out_err; |
f45827e8 | 1365 | |
e9be9d5e | 1366 | err = -EINVAL; |
b36a5780 | 1367 | if (ctx->nr == 0) { |
1784fbc2 | 1368 | if (!(fc->sb_flags & SB_SILENT)) |
1bd0a3ae | 1369 | pr_err("missing 'lowerdir'\n"); |
a9075cdb | 1370 | goto out_err; |
e9be9d5e MS |
1371 | } |
1372 | ||
b8e42a65 | 1373 | err = -ENOMEM; |
b36a5780 | 1374 | layers = kcalloc(ctx->nr + 1, sizeof(struct ovl_layer), GFP_KERNEL); |
b8e42a65 MS |
1375 | if (!layers) |
1376 | goto out_err; | |
1377 | ||
1378 | ofs->layers = layers; | |
1379 | /* Layer 0 is reserved for upper even if there's no upper */ | |
1380 | ofs->numlayer = 1; | |
1381 | ||
53a08cb9 | 1382 | sb->s_stack_depth = 0; |
cf9a6784 | 1383 | sb->s_maxbytes = MAX_LFS_FILESIZE; |
4d314f78 | 1384 | atomic_long_set(&ofs->last_ino, 1); |
4f119628 | 1385 | /* Assume underlying fs uses 32bit inodes unless proven otherwise */ |
53afcd31 | 1386 | if (ofs->config.xino != OVL_XINO_OFF) { |
0f831ec8 | 1387 | ofs->xino_mode = BITS_PER_LONG - 32; |
53afcd31 AG |
1388 | if (!ofs->xino_mode) { |
1389 | pr_warn("xino not supported on 32bit kernel, falling back to xino=off.\n"); | |
1390 | ofs->config.xino = OVL_XINO_OFF; | |
1391 | } | |
1392 | } | |
795939a9 | 1393 | |
146d62e5 AG |
1394 | /* alloc/destroy_inode needed for setting up traps in inode cache */ |
1395 | sb->s_op = &ovl_super_operations; | |
1396 | ||
ad204488 | 1397 | if (ofs->config.upperdir) { |
335d3fc5 SD |
1398 | struct super_block *upper_sb; |
1399 | ||
d7b49b10 | 1400 | err = -EINVAL; |
ad204488 | 1401 | if (!ofs->config.workdir) { |
1bd0a3ae | 1402 | pr_err("missing 'workdir'\n"); |
a9075cdb | 1403 | goto out_err; |
53a08cb9 | 1404 | } |
e9be9d5e | 1405 | |
b36a5780 | 1406 | err = ovl_get_upper(sb, ofs, &layers[0], &ctx->upper); |
53a08cb9 | 1407 | if (err) |
a9075cdb | 1408 | goto out_err; |
2cac0c00 | 1409 | |
335d3fc5 SD |
1410 | upper_sb = ovl_upper_mnt(ofs)->mnt_sb; |
1411 | if (!ovl_should_sync(ofs)) { | |
1412 | ofs->errseq = errseq_sample(&upper_sb->s_wb_err); | |
1413 | if (errseq_check(&upper_sb->s_wb_err, ofs->errseq)) { | |
1414 | err = -EIO; | |
1415 | pr_err("Cannot mount volatile when upperdir has an unseen error. Sync upperdir fs to clear state.\n"); | |
1416 | goto out_err; | |
1417 | } | |
1418 | } | |
1419 | ||
b36a5780 | 1420 | err = ovl_get_workdir(sb, ofs, &ctx->upper, &ctx->work); |
8ed61dc3 | 1421 | if (err) |
a9075cdb | 1422 | goto out_err; |
c6fe6254 | 1423 | |
ad204488 | 1424 | if (!ofs->workdir) |
1751e8a6 | 1425 | sb->s_flags |= SB_RDONLY; |
6e88256e | 1426 | |
335d3fc5 SD |
1427 | sb->s_stack_depth = upper_sb->s_stack_depth; |
1428 | sb->s_time_gran = upper_sb->s_time_gran; | |
e9be9d5e | 1429 | } |
b36a5780 | 1430 | oe = ovl_get_lowerstack(sb, ctx, ofs, layers); |
4155c10a MS |
1431 | err = PTR_ERR(oe); |
1432 | if (IS_ERR(oe)) | |
a9075cdb | 1433 | goto out_err; |
e9be9d5e | 1434 | |
71cbad7e | 1435 | /* If the upper fs is nonexistent, we mark overlayfs r/o too */ |
08f4c7c8 | 1436 | if (!ovl_upper_mnt(ofs)) |
1751e8a6 | 1437 | sb->s_flags |= SB_RDONLY; |
e9be9d5e | 1438 | |
b0504bfe AG |
1439 | if (!ovl_origin_uuid(ofs) && ofs->numfs > 1) { |
1440 | pr_warn("The uuid=off requires a single fs for lower and upper, falling back to uuid=null.\n"); | |
1441 | ofs->config.uuid = OVL_UUID_NULL; | |
d9544c1b AG |
1442 | } else if (ovl_has_fsid(ofs) && ovl_upper_mnt(ofs)) { |
1443 | /* Use per instance persistent uuid/fsid */ | |
1444 | ovl_init_uuid_xattr(sb, ofs, &ctx->upper); | |
5830fb6b PT |
1445 | } |
1446 | ||
470c1563 | 1447 | if (!ovl_force_readonly(ofs) && ofs->config.index) { |
b36a5780 | 1448 | err = ovl_get_indexdir(sb, ofs, oe, &ctx->upper); |
54fb347e | 1449 | if (err) |
4155c10a | 1450 | goto out_free_oe; |
6e88256e | 1451 | |
972d0093 | 1452 | /* Force r/o mount with no index dir */ |
20396365 | 1453 | if (!ofs->indexdir) |
1751e8a6 | 1454 | sb->s_flags |= SB_RDONLY; |
02bcd157 AG |
1455 | } |
1456 | ||
146d62e5 AG |
1457 | err = ovl_check_overlapping_layers(sb, ofs); |
1458 | if (err) | |
1459 | goto out_free_oe; | |
1460 | ||
972d0093 | 1461 | /* Show index=off in /proc/mounts for forced r/o mount */ |
f168f109 | 1462 | if (!ofs->indexdir) { |
ad204488 | 1463 | ofs->config.index = false; |
08f4c7c8 | 1464 | if (ovl_upper_mnt(ofs) && ofs->config.nfs_export) { |
1bd0a3ae | 1465 | pr_warn("NFS export requires an index dir, falling back to nfs_export=off.\n"); |
f168f109 AG |
1466 | ofs->config.nfs_export = false; |
1467 | } | |
1468 | } | |
02bcd157 | 1469 | |
d5791044 | 1470 | if (ofs->config.metacopy && ofs->config.nfs_export) { |
1bd0a3ae | 1471 | pr_warn("NFS export is not supported with metadata only copy up, falling back to nfs_export=off.\n"); |
d5791044 VG |
1472 | ofs->config.nfs_export = false; |
1473 | } | |
1474 | ||
16aac5ad AG |
1475 | /* |
1476 | * Support encoding decodable file handles with nfs_export=on | |
1477 | * and encoding non-decodable file handles with nfs_export=off | |
1478 | * if all layers support file handles. | |
1479 | */ | |
8383f174 AG |
1480 | if (ofs->config.nfs_export) |
1481 | sb->s_export_op = &ovl_export_operations; | |
16aac5ad AG |
1482 | else if (!ofs->nofh) |
1483 | sb->s_export_op = &ovl_export_fid_operations; | |
8383f174 | 1484 | |
51f8f3c4 KK |
1485 | /* Never override disk quota limits or use reserved space */ |
1486 | cap_lower(cred->cap_effective, CAP_SYS_RESOURCE); | |
1487 | ||
655042cc | 1488 | sb->s_magic = OVERLAYFS_SUPER_MAGIC; |
2d2f2d73 MS |
1489 | sb->s_xattr = ofs->config.userxattr ? ovl_user_xattr_handlers : |
1490 | ovl_trusted_xattr_handlers; | |
ad204488 | 1491 | sb->s_fs_info = ofs; |
2bc5e5e8 | 1492 | #ifdef CONFIG_FS_POSIX_ACL |
7c4d37c2 | 1493 | sb->s_flags |= SB_POSIXACL; |
2bc5e5e8 | 1494 | #endif |
18b44bc5 | 1495 | sb->s_iflags |= SB_I_SKIP_SYNC | SB_I_IMA_UNVERIFIABLE_SIGNATURE; |
2bc5e5e8 CB |
1496 | /* |
1497 | * Ensure that umask handling is done by the filesystems used | |
1498 | * for the the upper layer instead of overlayfs as that would | |
1499 | * lead to unexpected results. | |
1500 | */ | |
1501 | sb->s_iflags |= SB_I_NOUMASK; | |
655042cc | 1502 | |
c6fe6254 | 1503 | err = -ENOMEM; |
b36a5780 | 1504 | root_dentry = ovl_get_root(sb, ctx->upper.dentry, oe); |
e9be9d5e | 1505 | if (!root_dentry) |
4155c10a | 1506 | goto out_free_oe; |
e9be9d5e | 1507 | |
e9be9d5e | 1508 | sb->s_root = root_dentry; |
e9be9d5e MS |
1509 | |
1510 | return 0; | |
1511 | ||
4155c10a | 1512 | out_free_oe: |
163db0da | 1513 | ovl_free_entry(oe); |
4155c10a | 1514 | out_err: |
ad204488 | 1515 | ovl_free_fs(ofs); |
1784fbc2 | 1516 | sb->s_fs_info = NULL; |
e9be9d5e MS |
1517 | return err; |
1518 | } | |
1519 | ||
f01d0889 | 1520 | struct file_system_type ovl_fs_type = { |
1784fbc2 CB |
1521 | .owner = THIS_MODULE, |
1522 | .name = "overlay", | |
1523 | .init_fs_context = ovl_init_fs_context, | |
1524 | .parameters = ovl_parameter_spec, | |
1525 | .fs_flags = FS_USERNS_MOUNT, | |
1526 | .kill_sb = kill_anon_super, | |
e9be9d5e | 1527 | }; |
ef94b186 | 1528 | MODULE_ALIAS_FS("overlay"); |
e9be9d5e | 1529 | |
13cf199d AG |
1530 | static void ovl_inode_init_once(void *foo) |
1531 | { | |
1532 | struct ovl_inode *oi = foo; | |
1533 | ||
1534 | inode_init_once(&oi->vfs_inode); | |
1535 | } | |
1536 | ||
e9be9d5e MS |
1537 | static int __init ovl_init(void) |
1538 | { | |
13cf199d AG |
1539 | int err; |
1540 | ||
1541 | ovl_inode_cachep = kmem_cache_create("ovl_inode", | |
1542 | sizeof(struct ovl_inode), 0, | |
1543 | (SLAB_RECLAIM_ACCOUNT| | |
1544 | SLAB_MEM_SPREAD|SLAB_ACCOUNT), | |
1545 | ovl_inode_init_once); | |
1546 | if (ovl_inode_cachep == NULL) | |
1547 | return -ENOMEM; | |
1548 | ||
2406a307 JX |
1549 | err = ovl_aio_request_cache_init(); |
1550 | if (!err) { | |
1551 | err = register_filesystem(&ovl_fs_type); | |
1552 | if (!err) | |
1553 | return 0; | |
1554 | ||
1555 | ovl_aio_request_cache_destroy(); | |
1556 | } | |
1557 | kmem_cache_destroy(ovl_inode_cachep); | |
13cf199d AG |
1558 | |
1559 | return err; | |
e9be9d5e MS |
1560 | } |
1561 | ||
1562 | static void __exit ovl_exit(void) | |
1563 | { | |
1564 | unregister_filesystem(&ovl_fs_type); | |
13cf199d AG |
1565 | |
1566 | /* | |
1567 | * Make sure all delayed rcu free inodes are flushed before we | |
1568 | * destroy cache. | |
1569 | */ | |
1570 | rcu_barrier(); | |
1571 | kmem_cache_destroy(ovl_inode_cachep); | |
2406a307 | 1572 | ovl_aio_request_cache_destroy(); |
e9be9d5e MS |
1573 | } |
1574 | ||
1575 | module_init(ovl_init); | |
1576 | module_exit(ovl_exit); |