]>
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 | { |
c54719c9 AV |
115 | struct ovl_entry *oe; |
116 | struct ovl_path *lowerstack; | |
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 | ||
c54719c9 AV |
126 | oe = OVL_I_E(inode); |
127 | lowerstack = ovl_lowerstack(oe); | |
672e4268 | 128 | upper = ovl_i_dentry_upper(inode); |
bccece1e MS |
129 | if (upper) |
130 | ret = ovl_revalidate_real(upper, flags, weak); | |
131 | ||
5522c9c7 AG |
132 | for (i = 0; ret > 0 && i < ovl_numlower(oe); i++) |
133 | ret = ovl_revalidate_real(lowerstack[i].dentry, flags, weak); | |
134 | ||
7c03b5d4 MS |
135 | return ret; |
136 | } | |
137 | ||
3bb7df92 MS |
138 | static int ovl_dentry_revalidate(struct dentry *dentry, unsigned int flags) |
139 | { | |
140 | return ovl_dentry_revalidate_common(dentry, flags, false); | |
141 | } | |
142 | ||
143 | static int ovl_dentry_weak_revalidate(struct dentry *dentry, unsigned int flags) | |
144 | { | |
145 | return ovl_dentry_revalidate_common(dentry, flags, true); | |
146 | } | |
147 | ||
e9be9d5e | 148 | static const struct dentry_operations ovl_dentry_operations = { |
d101a125 | 149 | .d_real = ovl_d_real, |
7c03b5d4 MS |
150 | .d_revalidate = ovl_dentry_revalidate, |
151 | .d_weak_revalidate = ovl_dentry_weak_revalidate, | |
152 | }; | |
153 | ||
13cf199d AG |
154 | static struct kmem_cache *ovl_inode_cachep; |
155 | ||
156 | static struct inode *ovl_alloc_inode(struct super_block *sb) | |
157 | { | |
fd60b288 | 158 | struct ovl_inode *oi = alloc_inode_sb(sb, ovl_inode_cachep, GFP_KERNEL); |
13cf199d | 159 | |
b3885bd6 HN |
160 | if (!oi) |
161 | return NULL; | |
162 | ||
04a01ac7 | 163 | oi->cache = NULL; |
cf31c463 | 164 | oi->redirect = NULL; |
04a01ac7 | 165 | oi->version = 0; |
13c72075 | 166 | oi->flags = 0; |
09d8b586 | 167 | oi->__upperdentry = NULL; |
2b21da92 | 168 | oi->lowerdata_redirect = NULL; |
0af950f5 | 169 | oi->oe = NULL; |
a015dafc | 170 | mutex_init(&oi->lock); |
25b7713a | 171 | |
13cf199d AG |
172 | return &oi->vfs_inode; |
173 | } | |
174 | ||
0b269ded | 175 | static void ovl_free_inode(struct inode *inode) |
13cf199d | 176 | { |
0b269ded | 177 | struct ovl_inode *oi = OVL_I(inode); |
13cf199d | 178 | |
0b269ded | 179 | kfree(oi->redirect); |
d9e8319a | 180 | kfree(oi->oe); |
0b269ded AV |
181 | mutex_destroy(&oi->lock); |
182 | kmem_cache_free(ovl_inode_cachep, oi); | |
13cf199d AG |
183 | } |
184 | ||
185 | static void ovl_destroy_inode(struct inode *inode) | |
186 | { | |
09d8b586 MS |
187 | struct ovl_inode *oi = OVL_I(inode); |
188 | ||
189 | dput(oi->__upperdentry); | |
d9e8319a | 190 | ovl_stack_put(ovl_lowerstack(oi->oe), ovl_numlower(oi->oe)); |
2664bd08 VG |
191 | if (S_ISDIR(inode->i_mode)) |
192 | ovl_dir_cache_free(inode); | |
2b21da92 AG |
193 | else |
194 | kfree(oi->lowerdata_redirect); | |
13cf199d AG |
195 | } |
196 | ||
a9075cdb MS |
197 | static void ovl_put_super(struct super_block *sb) |
198 | { | |
f01d0889 | 199 | struct ovl_fs *ofs = OVL_FS(sb); |
a9075cdb | 200 | |
1784fbc2 CB |
201 | if (ofs) |
202 | ovl_free_fs(ofs); | |
a9075cdb MS |
203 | } |
204 | ||
e8d4bfe3 | 205 | /* Sync real dirty inodes in upper filesystem (if it exists) */ |
e593b2bf AG |
206 | static int ovl_sync_fs(struct super_block *sb, int wait) |
207 | { | |
f01d0889 | 208 | struct ovl_fs *ofs = OVL_FS(sb); |
e593b2bf AG |
209 | struct super_block *upper_sb; |
210 | int ret; | |
211 | ||
335d3fc5 SD |
212 | ret = ovl_sync_status(ofs); |
213 | /* | |
214 | * We have to always set the err, because the return value isn't | |
215 | * checked in syncfs, and instead indirectly return an error via | |
216 | * the sb's writeback errseq, which VFS inspects after this call. | |
217 | */ | |
218 | if (ret < 0) { | |
219 | errseq_set(&sb->s_wb_err, -EIO); | |
220 | return -EIO; | |
221 | } | |
222 | ||
223 | if (!ret) | |
224 | return ret; | |
e8d4bfe3 CX |
225 | |
226 | /* | |
32b1924b KK |
227 | * Not called for sync(2) call or an emergency sync (SB_I_SKIP_SYNC). |
228 | * All the super blocks will be iterated, including upper_sb. | |
e8d4bfe3 CX |
229 | * |
230 | * If this is a syncfs(2) call, then we do need to call | |
231 | * sync_filesystem() on upper_sb, but enough if we do it when being | |
232 | * called with wait == 1. | |
233 | */ | |
234 | if (!wait) | |
e593b2bf AG |
235 | return 0; |
236 | ||
08f4c7c8 | 237 | upper_sb = ovl_upper_mnt(ofs)->mnt_sb; |
e8d4bfe3 | 238 | |
e593b2bf | 239 | down_read(&upper_sb->s_umount); |
e8d4bfe3 | 240 | ret = sync_filesystem(upper_sb); |
e593b2bf | 241 | up_read(&upper_sb->s_umount); |
e8d4bfe3 | 242 | |
e593b2bf AG |
243 | return ret; |
244 | } | |
245 | ||
cc259639 AW |
246 | /** |
247 | * ovl_statfs | |
9c5dd803 | 248 | * @dentry: The dentry to query |
cc259639 AW |
249 | * @buf: The struct kstatfs to fill in with stats |
250 | * | |
251 | * Get the filesystem statistics. As writes always target the upper layer | |
4ebc5818 | 252 | * filesystem pass the statfs to the upper filesystem (if it exists) |
cc259639 AW |
253 | */ |
254 | static int ovl_statfs(struct dentry *dentry, struct kstatfs *buf) | |
255 | { | |
b0504bfe AG |
256 | struct super_block *sb = dentry->d_sb; |
257 | struct ovl_fs *ofs = OVL_FS(sb); | |
258 | struct dentry *root_dentry = sb->s_root; | |
cc259639 AW |
259 | struct path path; |
260 | int err; | |
261 | ||
4ebc5818 | 262 | ovl_path_real(root_dentry, &path); |
cc259639 AW |
263 | |
264 | err = vfs_statfs(&path, buf); | |
265 | if (!err) { | |
6b2d5fe4 | 266 | buf->f_namelen = ofs->namelen; |
cc259639 | 267 | buf->f_type = OVERLAYFS_SUPER_MAGIC; |
b0504bfe AG |
268 | if (ovl_has_fsid(ofs)) |
269 | buf->f_fsid = uuid_to_fsid(sb->s_uuid.b); | |
cc259639 AW |
270 | } |
271 | ||
272 | return err; | |
273 | } | |
274 | ||
e9be9d5e | 275 | static const struct super_operations ovl_super_operations = { |
13cf199d | 276 | .alloc_inode = ovl_alloc_inode, |
0b269ded | 277 | .free_inode = ovl_free_inode, |
13cf199d AG |
278 | .destroy_inode = ovl_destroy_inode, |
279 | .drop_inode = generic_delete_inode, | |
e9be9d5e | 280 | .put_super = ovl_put_super, |
e593b2bf | 281 | .sync_fs = ovl_sync_fs, |
cc259639 | 282 | .statfs = ovl_statfs, |
f45827e8 | 283 | .show_options = ovl_show_options, |
e9be9d5e MS |
284 | }; |
285 | ||
e9be9d5e | 286 | #define OVL_WORKDIR_NAME "work" |
02bcd157 | 287 | #define OVL_INDEXDIR_NAME "index" |
e9be9d5e | 288 | |
ad204488 | 289 | static struct dentry *ovl_workdir_create(struct ovl_fs *ofs, |
6b8aa129 | 290 | const char *name, bool persist) |
e9be9d5e | 291 | { |
ad204488 | 292 | struct inode *dir = ofs->workbasedir->d_inode; |
08f4c7c8 | 293 | struct vfsmount *mnt = ovl_upper_mnt(ofs); |
e9be9d5e MS |
294 | struct dentry *work; |
295 | int err; | |
296 | bool retried = false; | |
297 | ||
5955102c | 298 | inode_lock_nested(dir, I_MUTEX_PARENT); |
e9be9d5e | 299 | retry: |
22f289ce | 300 | work = ovl_lookup_upper(ofs, name, ofs->workbasedir, strlen(name)); |
e9be9d5e MS |
301 | |
302 | if (!IS_ERR(work)) { | |
c11b9fdd MS |
303 | struct iattr attr = { |
304 | .ia_valid = ATTR_MODE, | |
32a3d848 | 305 | .ia_mode = S_IFDIR | 0, |
c11b9fdd | 306 | }; |
e9be9d5e MS |
307 | |
308 | if (work->d_inode) { | |
309 | err = -EEXIST; | |
310 | if (retried) | |
311 | goto out_dput; | |
312 | ||
6b8aa129 AG |
313 | if (persist) |
314 | goto out_unlock; | |
315 | ||
e9be9d5e | 316 | retried = true; |
576bb263 | 317 | err = ovl_workdir_cleanup(ofs, dir, mnt, work, 0); |
e9be9d5e | 318 | dput(work); |
235ce9ed AG |
319 | if (err == -EINVAL) { |
320 | work = ERR_PTR(err); | |
321 | goto out_unlock; | |
322 | } | |
e9be9d5e MS |
323 | goto retry; |
324 | } | |
325 | ||
576bb263 | 326 | err = ovl_mkdir_real(ofs, dir, &work, attr.ia_mode); |
1f5573cf MS |
327 | if (err) |
328 | goto out_dput; | |
329 | ||
330 | /* Weird filesystem returning with hashed negative (kernfs)? */ | |
331 | err = -EINVAL; | |
332 | if (d_really_is_negative(work)) | |
333 | goto out_dput; | |
c11b9fdd | 334 | |
cb348edb MS |
335 | /* |
336 | * Try to remove POSIX ACL xattrs from workdir. We are good if: | |
337 | * | |
338 | * a) success (there was a POSIX ACL xattr and was removed) | |
339 | * b) -ENODATA (there was no POSIX ACL xattr) | |
340 | * c) -EOPNOTSUPP (POSIX ACL xattrs are not supported) | |
341 | * | |
342 | * There are various other error values that could effectively | |
343 | * mean that the xattr doesn't exist (e.g. -ERANGE is returned | |
344 | * if the xattr name is too long), but the set of filesystems | |
345 | * allowed as upper are limited to "normal" ones, where checking | |
346 | * for the above two errors is sufficient. | |
347 | */ | |
31acceb9 | 348 | err = ovl_do_remove_acl(ofs, work, XATTR_NAME_POSIX_ACL_DEFAULT); |
e1ff3dd1 | 349 | if (err && err != -ENODATA && err != -EOPNOTSUPP) |
c11b9fdd MS |
350 | goto out_dput; |
351 | ||
31acceb9 | 352 | err = ovl_do_remove_acl(ofs, work, XATTR_NAME_POSIX_ACL_ACCESS); |
e1ff3dd1 | 353 | if (err && err != -ENODATA && err != -EOPNOTSUPP) |
c11b9fdd MS |
354 | goto out_dput; |
355 | ||
356 | /* Clear any inherited mode bits */ | |
357 | inode_lock(work->d_inode); | |
a15506ea | 358 | err = ovl_do_notify_change(ofs, work, &attr); |
c11b9fdd MS |
359 | inode_unlock(work->d_inode); |
360 | if (err) | |
361 | goto out_dput; | |
6b8aa129 AG |
362 | } else { |
363 | err = PTR_ERR(work); | |
364 | goto out_err; | |
e9be9d5e MS |
365 | } |
366 | out_unlock: | |
2068cf7d | 367 | inode_unlock(dir); |
e9be9d5e MS |
368 | return work; |
369 | ||
370 | out_dput: | |
371 | dput(work); | |
6b8aa129 | 372 | out_err: |
1bd0a3ae | 373 | pr_warn("failed to create directory %s/%s (errno: %i); mounting read-only\n", |
ad204488 | 374 | ofs->config.workdir, name, -err); |
6b8aa129 | 375 | work = NULL; |
e9be9d5e MS |
376 | goto out_unlock; |
377 | } | |
378 | ||
2d343087 | 379 | static int ovl_check_namelen(const struct path *path, struct ovl_fs *ofs, |
6b2d5fe4 | 380 | const char *name) |
ab508822 | 381 | { |
ab508822 | 382 | struct kstatfs statfs; |
6b2d5fe4 MS |
383 | int err = vfs_statfs(path, &statfs); |
384 | ||
385 | if (err) | |
1bd0a3ae | 386 | pr_err("statfs failed on '%s'\n", name); |
6b2d5fe4 MS |
387 | else |
388 | ofs->namelen = max(ofs->namelen, statfs.f_namelen); | |
389 | ||
390 | return err; | |
391 | } | |
392 | ||
393 | static int ovl_lower_dir(const char *name, struct path *path, | |
f4288844 | 394 | struct ovl_fs *ofs, int *stack_depth) |
6b2d5fe4 | 395 | { |
e487d889 | 396 | int fh_type; |
6b2d5fe4 | 397 | int err; |
ab508822 | 398 | |
6b2d5fe4 MS |
399 | err = ovl_check_namelen(path, ofs, name); |
400 | if (err) | |
b8e42a65 | 401 | return err; |
6b2d5fe4 | 402 | |
ab508822 MS |
403 | *stack_depth = max(*stack_depth, path->mnt->mnt_sb->s_stack_depth); |
404 | ||
02bcd157 | 405 | /* |
f168f109 AG |
406 | * The inodes index feature and NFS export need to encode and decode |
407 | * file handles, so they require that all layers support them. | |
02bcd157 | 408 | */ |
e487d889 | 409 | fh_type = ovl_can_decode_fh(path->dentry->d_sb); |
f168f109 | 410 | if ((ofs->config.nfs_export || |
e487d889 | 411 | (ofs->config.index && ofs->config.upperdir)) && !fh_type) { |
02bcd157 | 412 | ofs->config.index = false; |
f168f109 | 413 | ofs->config.nfs_export = false; |
1bd0a3ae | 414 | pr_warn("fs on '%s' does not support file handles, falling back to index=off,nfs_export=off.\n", |
f168f109 | 415 | name); |
02bcd157 | 416 | } |
16aac5ad | 417 | ofs->nofh |= !fh_type; |
b0e0f697 AG |
418 | /* |
419 | * Decoding origin file handle is required for persistent st_ino. | |
420 | * Without persistent st_ino, xino=auto falls back to xino=off. | |
421 | */ | |
422 | if (ofs->config.xino == OVL_XINO_AUTO && | |
423 | ofs->config.upperdir && !fh_type) { | |
424 | ofs->config.xino = OVL_XINO_OFF; | |
425 | pr_warn("fs on '%s' does not support file handles, falling back to xino=off.\n", | |
426 | name); | |
427 | } | |
02bcd157 | 428 | |
e487d889 AG |
429 | /* Check if lower fs has 32bit inode numbers */ |
430 | if (fh_type != FILEID_INO32_GEN) | |
0f831ec8 | 431 | ofs->xino_mode = -1; |
e487d889 | 432 | |
ab508822 | 433 | return 0; |
ab508822 MS |
434 | } |
435 | ||
e9be9d5e MS |
436 | /* Workdir should not be subdir of upperdir and vice versa */ |
437 | static bool ovl_workdir_ok(struct dentry *workdir, struct dentry *upperdir) | |
438 | { | |
439 | bool ok = false; | |
440 | ||
441 | if (workdir != upperdir) { | |
a8b00268 AV |
442 | struct dentry *trap = lock_rename(workdir, upperdir); |
443 | if (!IS_ERR(trap)) | |
444 | unlock_rename(workdir, upperdir); | |
445 | ok = (trap == NULL); | |
e9be9d5e MS |
446 | } |
447 | return ok; | |
448 | } | |
449 | ||
146d62e5 AG |
450 | static int ovl_setup_trap(struct super_block *sb, struct dentry *dir, |
451 | struct inode **ptrap, const char *name) | |
452 | { | |
453 | struct inode *trap; | |
454 | int err; | |
455 | ||
456 | trap = ovl_get_trap_inode(sb, dir); | |
1dac6f5b AB |
457 | err = PTR_ERR_OR_ZERO(trap); |
458 | if (err) { | |
146d62e5 | 459 | if (err == -ELOOP) |
1bd0a3ae | 460 | pr_err("conflicting %s path\n", name); |
146d62e5 AG |
461 | return err; |
462 | } | |
463 | ||
464 | *ptrap = trap; | |
465 | return 0; | |
466 | } | |
467 | ||
0be0bfd2 AG |
468 | /* |
469 | * Determine how we treat concurrent use of upperdir/workdir based on the | |
470 | * index feature. This is papering over mount leaks of container runtimes, | |
471 | * for example, an old overlay mount is leaked and now its upperdir is | |
472 | * attempted to be used as a lower layer in a new overlay mount. | |
473 | */ | |
474 | static int ovl_report_in_use(struct ovl_fs *ofs, const char *name) | |
475 | { | |
476 | if (ofs->config.index) { | |
1bd0a3ae | 477 | 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 |
478 | name); |
479 | return -EBUSY; | |
480 | } else { | |
1bd0a3ae | 481 | 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 |
482 | name); |
483 | return 0; | |
484 | } | |
485 | } | |
486 | ||
146d62e5 | 487 | static int ovl_get_upper(struct super_block *sb, struct ovl_fs *ofs, |
b36a5780 CB |
488 | struct ovl_layer *upper_layer, |
489 | const struct path *upperpath) | |
6ee8acf0 | 490 | { |
5064975e | 491 | struct vfsmount *upper_mnt; |
6ee8acf0 MS |
492 | int err; |
493 | ||
e21a6c57 AG |
494 | /* Upperdir path should not be r/o */ |
495 | if (__mnt_is_readonly(upperpath->mnt)) { | |
1bd0a3ae | 496 | pr_err("upper fs is r/o, try multi-lower layers mount\n"); |
6ee8acf0 MS |
497 | err = -EINVAL; |
498 | goto out; | |
499 | } | |
500 | ||
ad204488 | 501 | err = ovl_check_namelen(upperpath, ofs, ofs->config.upperdir); |
6ee8acf0 MS |
502 | if (err) |
503 | goto out; | |
504 | ||
b8e42a65 | 505 | err = ovl_setup_trap(sb, upperpath->dentry, &upper_layer->trap, |
146d62e5 AG |
506 | "upperdir"); |
507 | if (err) | |
508 | goto out; | |
509 | ||
5064975e MS |
510 | upper_mnt = clone_private_mount(upperpath); |
511 | err = PTR_ERR(upper_mnt); | |
512 | if (IS_ERR(upper_mnt)) { | |
1bd0a3ae | 513 | pr_err("failed to clone upperpath\n"); |
5064975e MS |
514 | goto out; |
515 | } | |
516 | ||
517 | /* Don't inherit atime flags */ | |
518 | upper_mnt->mnt_flags &= ~(MNT_NOATIME | MNT_NODIRATIME | MNT_RELATIME); | |
b8e42a65 MS |
519 | upper_layer->mnt = upper_mnt; |
520 | upper_layer->idx = 0; | |
521 | upper_layer->fsid = 0; | |
8c25741a | 522 | |
654255fa JX |
523 | /* |
524 | * Inherit SB_NOSEC flag from upperdir. | |
525 | * | |
526 | * This optimization changes behavior when a security related attribute | |
527 | * (suid/sgid/security.*) is changed on an underlying layer. This is | |
528 | * okay because we don't yet have guarantees in that case, but it will | |
529 | * need careful treatment once we want to honour changes to underlying | |
530 | * filesystems. | |
531 | */ | |
532 | if (upper_mnt->mnt_sb->s_flags & SB_NOSEC) | |
533 | sb->s_flags |= SB_NOSEC; | |
534 | ||
08f4c7c8 | 535 | if (ovl_inuse_trylock(ovl_upper_mnt(ofs)->mnt_root)) { |
8c25741a | 536 | ofs->upperdir_locked = true; |
8c25741a | 537 | } else { |
0be0bfd2 AG |
538 | err = ovl_report_in_use(ofs, "upperdir"); |
539 | if (err) | |
540 | goto out; | |
8c25741a MS |
541 | } |
542 | ||
6ee8acf0 MS |
543 | err = 0; |
544 | out: | |
545 | return err; | |
546 | } | |
547 | ||
cad218ab AG |
548 | /* |
549 | * Returns 1 if RENAME_WHITEOUT is supported, 0 if not supported and | |
550 | * negative values if error is encountered. | |
551 | */ | |
576bb263 | 552 | static int ovl_check_rename_whiteout(struct ovl_fs *ofs) |
cad218ab | 553 | { |
576bb263 | 554 | struct dentry *workdir = ofs->workdir; |
cad218ab AG |
555 | struct inode *dir = d_inode(workdir); |
556 | struct dentry *temp; | |
557 | struct dentry *dest; | |
558 | struct dentry *whiteout; | |
559 | struct name_snapshot name; | |
560 | int err; | |
561 | ||
562 | inode_lock_nested(dir, I_MUTEX_PARENT); | |
563 | ||
576bb263 | 564 | temp = ovl_create_temp(ofs, workdir, OVL_CATTR(S_IFREG | 0)); |
cad218ab AG |
565 | err = PTR_ERR(temp); |
566 | if (IS_ERR(temp)) | |
567 | goto out_unlock; | |
568 | ||
576bb263 | 569 | dest = ovl_lookup_temp(ofs, workdir); |
cad218ab AG |
570 | err = PTR_ERR(dest); |
571 | if (IS_ERR(dest)) { | |
572 | dput(temp); | |
573 | goto out_unlock; | |
574 | } | |
575 | ||
576 | /* Name is inline and stable - using snapshot as a copy helper */ | |
577 | take_dentry_name_snapshot(&name, temp); | |
576bb263 | 578 | err = ovl_do_rename(ofs, dir, temp, dir, dest, RENAME_WHITEOUT); |
cad218ab AG |
579 | if (err) { |
580 | if (err == -EINVAL) | |
581 | err = 0; | |
582 | goto cleanup_temp; | |
583 | } | |
584 | ||
22f289ce | 585 | whiteout = ovl_lookup_upper(ofs, name.name.name, workdir, name.name.len); |
cad218ab AG |
586 | err = PTR_ERR(whiteout); |
587 | if (IS_ERR(whiteout)) | |
588 | goto cleanup_temp; | |
589 | ||
bc8df7a3 | 590 | err = ovl_upper_is_whiteout(ofs, whiteout); |
cad218ab AG |
591 | |
592 | /* Best effort cleanup of whiteout and temp file */ | |
593 | if (err) | |
576bb263 | 594 | ovl_cleanup(ofs, dir, whiteout); |
cad218ab AG |
595 | dput(whiteout); |
596 | ||
597 | cleanup_temp: | |
576bb263 | 598 | ovl_cleanup(ofs, dir, temp); |
cad218ab AG |
599 | release_dentry_name_snapshot(&name); |
600 | dput(temp); | |
601 | dput(dest); | |
602 | ||
603 | out_unlock: | |
604 | inode_unlock(dir); | |
605 | ||
606 | return err; | |
607 | } | |
608 | ||
576bb263 CB |
609 | static struct dentry *ovl_lookup_or_create(struct ovl_fs *ofs, |
610 | struct dentry *parent, | |
c86243b0 VG |
611 | const char *name, umode_t mode) |
612 | { | |
613 | size_t len = strlen(name); | |
614 | struct dentry *child; | |
615 | ||
616 | inode_lock_nested(parent->d_inode, I_MUTEX_PARENT); | |
22f289ce | 617 | child = ovl_lookup_upper(ofs, name, parent, len); |
c86243b0 | 618 | if (!IS_ERR(child) && !child->d_inode) |
576bb263 | 619 | child = ovl_create_real(ofs, parent->d_inode, child, |
c86243b0 VG |
620 | OVL_CATTR(mode)); |
621 | inode_unlock(parent->d_inode); | |
622 | dput(parent); | |
623 | ||
624 | return child; | |
625 | } | |
626 | ||
627 | /* | |
628 | * Creates $workdir/work/incompat/volatile/dirty file if it is not already | |
629 | * present. | |
630 | */ | |
631 | static int ovl_create_volatile_dirty(struct ovl_fs *ofs) | |
632 | { | |
633 | unsigned int ctr; | |
634 | struct dentry *d = dget(ofs->workbasedir); | |
635 | static const char *const volatile_path[] = { | |
636 | OVL_WORKDIR_NAME, "incompat", "volatile", "dirty" | |
637 | }; | |
638 | const char *const *name = volatile_path; | |
639 | ||
640 | for (ctr = ARRAY_SIZE(volatile_path); ctr; ctr--, name++) { | |
576bb263 | 641 | d = ovl_lookup_or_create(ofs, d, *name, ctr > 1 ? S_IFDIR : S_IFREG); |
c86243b0 VG |
642 | if (IS_ERR(d)) |
643 | return PTR_ERR(d); | |
644 | } | |
645 | dput(d); | |
646 | return 0; | |
647 | } | |
648 | ||
146d62e5 | 649 | static int ovl_make_workdir(struct super_block *sb, struct ovl_fs *ofs, |
2d343087 | 650 | const struct path *workpath) |
8ed61dc3 | 651 | { |
08f4c7c8 | 652 | struct vfsmount *mnt = ovl_upper_mnt(ofs); |
2b1a7746 MS |
653 | struct dentry *workdir; |
654 | struct file *tmpfile; | |
d80172c2 AG |
655 | bool rename_whiteout; |
656 | bool d_type; | |
e487d889 | 657 | int fh_type; |
8ed61dc3 MS |
658 | int err; |
659 | ||
2ba9d57e AG |
660 | err = mnt_want_write(mnt); |
661 | if (err) | |
662 | return err; | |
663 | ||
235ce9ed AG |
664 | workdir = ovl_workdir_create(ofs, OVL_WORKDIR_NAME, false); |
665 | err = PTR_ERR(workdir); | |
666 | if (IS_ERR_OR_NULL(workdir)) | |
2ba9d57e | 667 | goto out; |
8ed61dc3 | 668 | |
235ce9ed AG |
669 | ofs->workdir = workdir; |
670 | ||
146d62e5 AG |
671 | err = ovl_setup_trap(sb, ofs->workdir, &ofs->workdir_trap, "workdir"); |
672 | if (err) | |
673 | goto out; | |
674 | ||
8ed61dc3 MS |
675 | /* |
676 | * Upper should support d_type, else whiteouts are visible. Given | |
677 | * workdir and upper are on same fs, we can do iterate_dir() on | |
678 | * workdir. This check requires successful creation of workdir in | |
679 | * previous step. | |
680 | */ | |
681 | err = ovl_check_d_type_supported(workpath); | |
682 | if (err < 0) | |
2ba9d57e | 683 | goto out; |
8ed61dc3 | 684 | |
d80172c2 AG |
685 | d_type = err; |
686 | if (!d_type) | |
1bd0a3ae | 687 | pr_warn("upper fs needs to support d_type.\n"); |
8ed61dc3 MS |
688 | |
689 | /* Check if upper/work fs supports O_TMPFILE */ | |
2b1a7746 MS |
690 | tmpfile = ovl_do_tmpfile(ofs, ofs->workdir, S_IFREG | 0); |
691 | ofs->tmpfile = !IS_ERR(tmpfile); | |
ad204488 | 692 | if (ofs->tmpfile) |
2b1a7746 | 693 | fput(tmpfile); |
8ed61dc3 | 694 | else |
1bd0a3ae | 695 | pr_warn("upper fs does not support tmpfile.\n"); |
8ed61dc3 | 696 | |
cad218ab AG |
697 | |
698 | /* Check if upper/work fs supports RENAME_WHITEOUT */ | |
576bb263 | 699 | err = ovl_check_rename_whiteout(ofs); |
cad218ab AG |
700 | if (err < 0) |
701 | goto out; | |
702 | ||
d80172c2 AG |
703 | rename_whiteout = err; |
704 | if (!rename_whiteout) | |
cad218ab AG |
705 | pr_warn("upper fs does not support RENAME_WHITEOUT.\n"); |
706 | ||
8ed61dc3 | 707 | /* |
2d2f2d73 | 708 | * Check if upper/work fs supports (trusted|user).overlay.* xattr |
8ed61dc3 | 709 | */ |
c914c0e2 | 710 | err = ovl_setxattr(ofs, ofs->workdir, OVL_XATTR_OPAQUE, "0", 1); |
8ed61dc3 | 711 | if (err) { |
b10b85fe | 712 | pr_warn("failed to set xattr on upper\n"); |
ad204488 | 713 | ofs->noxattr = true; |
af5f2396 AG |
714 | if (ovl_redirect_follow(ofs)) { |
715 | ofs->config.redirect_mode = OVL_REDIRECT_NOFOLLOW; | |
716 | pr_warn("...falling back to redirect_dir=nofollow.\n"); | |
717 | } | |
718 | if (ofs->config.metacopy) { | |
b0e0f697 | 719 | ofs->config.metacopy = false; |
af5f2396 AG |
720 | pr_warn("...falling back to metacopy=off.\n"); |
721 | } | |
722 | if (ofs->config.index) { | |
723 | ofs->config.index = false; | |
724 | pr_warn("...falling back to index=off.\n"); | |
b0e0f697 | 725 | } |
d9544c1b AG |
726 | if (ovl_has_fsid(ofs)) { |
727 | ofs->config.uuid = OVL_UUID_NULL; | |
728 | pr_warn("...falling back to uuid=null.\n"); | |
729 | } | |
b0e0f697 AG |
730 | /* |
731 | * xattr support is required for persistent st_ino. | |
732 | * Without persistent st_ino, xino=auto falls back to xino=off. | |
733 | */ | |
734 | if (ofs->config.xino == OVL_XINO_AUTO) { | |
735 | ofs->config.xino = OVL_XINO_OFF; | |
b10b85fe | 736 | pr_warn("...falling back to xino=off.\n"); |
b0e0f697 | 737 | } |
b10b85fe MS |
738 | if (err == -EPERM && !ofs->config.userxattr) |
739 | pr_info("try mounting with 'userxattr' option\n"); | |
2ba9d57e | 740 | err = 0; |
8ed61dc3 | 741 | } else { |
c914c0e2 | 742 | ovl_removexattr(ofs, ofs->workdir, OVL_XATTR_OPAQUE); |
8ed61dc3 MS |
743 | } |
744 | ||
d80172c2 AG |
745 | /* |
746 | * We allowed sub-optimal upper fs configuration and don't want to break | |
747 | * users over kernel upgrade, but we never allowed remote upper fs, so | |
748 | * we can enforce strict requirements for remote upper fs. | |
749 | */ | |
750 | if (ovl_dentry_remote(ofs->workdir) && | |
751 | (!d_type || !rename_whiteout || ofs->noxattr)) { | |
752 | pr_err("upper fs missing required features.\n"); | |
753 | err = -EINVAL; | |
754 | goto out; | |
755 | } | |
756 | ||
c86243b0 VG |
757 | /* |
758 | * For volatile mount, create a incompat/volatile/dirty file to keep | |
759 | * track of it. | |
760 | */ | |
761 | if (ofs->config.ovl_volatile) { | |
762 | err = ovl_create_volatile_dirty(ofs); | |
763 | if (err < 0) { | |
764 | pr_err("Failed to create volatile/dirty file.\n"); | |
765 | goto out; | |
766 | } | |
767 | } | |
768 | ||
8ed61dc3 | 769 | /* Check if upper/work fs supports file handles */ |
e487d889 AG |
770 | fh_type = ovl_can_decode_fh(ofs->workdir->d_sb); |
771 | if (ofs->config.index && !fh_type) { | |
ad204488 | 772 | ofs->config.index = false; |
1bd0a3ae | 773 | pr_warn("upper fs does not support file handles, falling back to index=off.\n"); |
8ed61dc3 | 774 | } |
16aac5ad | 775 | ofs->nofh |= !fh_type; |
8ed61dc3 | 776 | |
e487d889 AG |
777 | /* Check if upper fs has 32bit inode numbers */ |
778 | if (fh_type != FILEID_INO32_GEN) | |
0f831ec8 | 779 | ofs->xino_mode = -1; |
e487d889 | 780 | |
f168f109 AG |
781 | /* NFS export of r/w mount depends on index */ |
782 | if (ofs->config.nfs_export && !ofs->config.index) { | |
1bd0a3ae | 783 | pr_warn("NFS export requires \"index=on\", falling back to nfs_export=off.\n"); |
f168f109 AG |
784 | ofs->config.nfs_export = false; |
785 | } | |
2ba9d57e AG |
786 | out: |
787 | mnt_drop_write(mnt); | |
788 | return err; | |
8ed61dc3 MS |
789 | } |
790 | ||
146d62e5 | 791 | static int ovl_get_workdir(struct super_block *sb, struct ovl_fs *ofs, |
b36a5780 CB |
792 | const struct path *upperpath, |
793 | const struct path *workpath) | |
520d7c86 MS |
794 | { |
795 | int err; | |
520d7c86 MS |
796 | |
797 | err = -EINVAL; | |
b36a5780 | 798 | if (upperpath->mnt != workpath->mnt) { |
1bd0a3ae | 799 | pr_err("workdir and upperdir must reside under the same mount\n"); |
b36a5780 | 800 | return err; |
520d7c86 | 801 | } |
b36a5780 | 802 | if (!ovl_workdir_ok(workpath->dentry, upperpath->dentry)) { |
1bd0a3ae | 803 | pr_err("workdir and upperdir must be separate subtrees\n"); |
b36a5780 | 804 | return err; |
520d7c86 MS |
805 | } |
806 | ||
b36a5780 | 807 | ofs->workbasedir = dget(workpath->dentry); |
8c25741a | 808 | |
8c25741a | 809 | if (ovl_inuse_trylock(ofs->workbasedir)) { |
ad204488 | 810 | ofs->workdir_locked = true; |
520d7c86 | 811 | } else { |
0be0bfd2 AG |
812 | err = ovl_report_in_use(ofs, "workdir"); |
813 | if (err) | |
b36a5780 | 814 | return err; |
520d7c86 MS |
815 | } |
816 | ||
0be0bfd2 AG |
817 | err = ovl_setup_trap(sb, ofs->workbasedir, &ofs->workbasedir_trap, |
818 | "workdir"); | |
819 | if (err) | |
b36a5780 | 820 | return err; |
bca44b52 | 821 | |
b36a5780 | 822 | return ovl_make_workdir(sb, ofs, workpath); |
520d7c86 MS |
823 | } |
824 | ||
146d62e5 | 825 | static int ovl_get_indexdir(struct super_block *sb, struct ovl_fs *ofs, |
2d343087 | 826 | struct ovl_entry *oe, const struct path *upperpath) |
f7e3a7d9 | 827 | { |
08f4c7c8 | 828 | struct vfsmount *mnt = ovl_upper_mnt(ofs); |
235ce9ed | 829 | struct dentry *indexdir; |
5b02bfc1 AG |
830 | struct dentry *origin = ovl_lowerstack(oe)->dentry; |
831 | const struct ovl_fh *fh; | |
f7e3a7d9 MS |
832 | int err; |
833 | ||
5b02bfc1 AG |
834 | fh = ovl_get_origin_fh(ofs, origin); |
835 | if (IS_ERR(fh)) | |
836 | return PTR_ERR(fh); | |
837 | ||
2ba9d57e AG |
838 | err = mnt_want_write(mnt); |
839 | if (err) | |
5b02bfc1 | 840 | goto out_free_fh; |
2ba9d57e | 841 | |
f7e3a7d9 | 842 | /* Verify lower root is upper root origin */ |
5b02bfc1 | 843 | err = ovl_verify_origin_fh(ofs, upperpath->dentry, fh, true); |
f7e3a7d9 | 844 | if (err) { |
1bd0a3ae | 845 | pr_err("failed to verify upper root origin\n"); |
f7e3a7d9 MS |
846 | goto out; |
847 | } | |
848 | ||
470c1563 AG |
849 | /* index dir will act also as workdir */ |
850 | iput(ofs->workdir_trap); | |
851 | ofs->workdir_trap = NULL; | |
852 | dput(ofs->workdir); | |
853 | ofs->workdir = NULL; | |
235ce9ed AG |
854 | indexdir = ovl_workdir_create(ofs, OVL_INDEXDIR_NAME, true); |
855 | if (IS_ERR(indexdir)) { | |
856 | err = PTR_ERR(indexdir); | |
857 | } else if (indexdir) { | |
02d70090 AG |
858 | ofs->workdir = indexdir; |
859 | err = ovl_setup_trap(sb, indexdir, &ofs->workdir_trap, | |
146d62e5 AG |
860 | "indexdir"); |
861 | if (err) | |
862 | goto out; | |
863 | ||
ad1d615c AG |
864 | /* |
865 | * Verify upper root is exclusively associated with index dir. | |
2d2f2d73 | 866 | * Older kernels stored upper fh in ".overlay.origin" |
ad1d615c AG |
867 | * xattr. If that xattr exists, verify that it is a match to |
868 | * upper dir file handle. In any case, verify or set xattr | |
2d2f2d73 | 869 | * ".overlay.upper" to indicate that index may have |
ad1d615c AG |
870 | * directory entries. |
871 | */ | |
02d70090 AG |
872 | if (ovl_check_origin_xattr(ofs, indexdir)) { |
873 | err = ovl_verify_origin_xattr(ofs, indexdir, | |
5b02bfc1 AG |
874 | OVL_XATTR_ORIGIN, |
875 | upperpath->dentry, true, | |
876 | false); | |
ad1d615c | 877 | if (err) |
1bd0a3ae | 878 | pr_err("failed to verify index dir 'origin' xattr\n"); |
ad1d615c | 879 | } |
02d70090 | 880 | err = ovl_verify_upper(ofs, indexdir, upperpath->dentry, true); |
f7e3a7d9 | 881 | if (err) |
1bd0a3ae | 882 | pr_err("failed to verify index dir 'upper' xattr\n"); |
f7e3a7d9 MS |
883 | |
884 | /* Cleanup bad/stale/orphan index entries */ | |
885 | if (!err) | |
1eff1a1d | 886 | err = ovl_indexdir_cleanup(ofs); |
f7e3a7d9 | 887 | } |
02d70090 | 888 | if (err || !indexdir) |
1bd0a3ae | 889 | pr_warn("try deleting index dir or mounting with '-o index=off' to disable inodes index.\n"); |
f7e3a7d9 MS |
890 | |
891 | out: | |
2ba9d57e | 892 | mnt_drop_write(mnt); |
5b02bfc1 AG |
893 | out_free_fh: |
894 | kfree(fh); | |
f7e3a7d9 MS |
895 | return err; |
896 | } | |
897 | ||
9df085f3 AG |
898 | static bool ovl_lower_uuid_ok(struct ovl_fs *ofs, const uuid_t *uuid) |
899 | { | |
900 | unsigned int i; | |
901 | ||
08f4c7c8 | 902 | if (!ofs->config.nfs_export && !ovl_upper_mnt(ofs)) |
9df085f3 AG |
903 | return true; |
904 | ||
a888db31 AG |
905 | /* |
906 | * We allow using single lower with null uuid for index and nfs_export | |
907 | * for example to support those features with single lower squashfs. | |
908 | * To avoid regressions in setups of overlay with re-formatted lower | |
909 | * squashfs, do not allow decoding origin with lower null uuid unless | |
910 | * user opted-in to one of the new features that require following the | |
911 | * lower inode of non-dir upper. | |
912 | */ | |
ca45275c | 913 | if (ovl_allow_offline_changes(ofs) && uuid_is_null(uuid)) |
a888db31 AG |
914 | return false; |
915 | ||
1b81dddd | 916 | for (i = 0; i < ofs->numfs; i++) { |
9df085f3 AG |
917 | /* |
918 | * We use uuid to associate an overlay lower file handle with a | |
919 | * lower layer, so we can accept lower fs with null uuid as long | |
920 | * as all lower layers with null uuid are on the same fs. | |
7e63c87f AG |
921 | * if we detect multiple lower fs with the same uuid, we |
922 | * disable lower file handle decoding on all of them. | |
9df085f3 | 923 | */ |
1b81dddd AG |
924 | if (ofs->fs[i].is_lower && |
925 | uuid_equal(&ofs->fs[i].sb->s_uuid, uuid)) { | |
07f1e596 | 926 | ofs->fs[i].bad_uuid = true; |
9df085f3 | 927 | return false; |
7e63c87f | 928 | } |
9df085f3 AG |
929 | } |
930 | return true; | |
931 | } | |
932 | ||
5148626b | 933 | /* Get a unique fsid for the layer */ |
9df085f3 | 934 | static int ovl_get_fsid(struct ovl_fs *ofs, const struct path *path) |
5148626b | 935 | { |
9df085f3 | 936 | struct super_block *sb = path->mnt->mnt_sb; |
5148626b AG |
937 | unsigned int i; |
938 | dev_t dev; | |
939 | int err; | |
7e63c87f | 940 | bool bad_uuid = false; |
b0e0f697 | 941 | bool warn = false; |
5148626b | 942 | |
07f1e596 AG |
943 | for (i = 0; i < ofs->numfs; i++) { |
944 | if (ofs->fs[i].sb == sb) | |
945 | return i; | |
5148626b AG |
946 | } |
947 | ||
9df085f3 | 948 | if (!ovl_lower_uuid_ok(ofs, &sb->s_uuid)) { |
7e63c87f | 949 | bad_uuid = true; |
b0e0f697 AG |
950 | if (ofs->config.xino == OVL_XINO_AUTO) { |
951 | ofs->config.xino = OVL_XINO_OFF; | |
952 | warn = true; | |
953 | } | |
7e63c87f AG |
954 | if (ofs->config.index || ofs->config.nfs_export) { |
955 | ofs->config.index = false; | |
956 | ofs->config.nfs_export = false; | |
b0e0f697 AG |
957 | warn = true; |
958 | } | |
959 | if (warn) { | |
960 | pr_warn("%s uuid detected in lower fs '%pd2', falling back to xino=%s,index=off,nfs_export=off.\n", | |
7e63c87f AG |
961 | uuid_is_null(&sb->s_uuid) ? "null" : |
962 | "conflicting", | |
dcb399de | 963 | path->dentry, ovl_xino_mode(&ofs->config)); |
7e63c87f | 964 | } |
9df085f3 AG |
965 | } |
966 | ||
5148626b AG |
967 | err = get_anon_bdev(&dev); |
968 | if (err) { | |
1bd0a3ae | 969 | pr_err("failed to get anonymous bdev for lowerpath\n"); |
5148626b AG |
970 | return err; |
971 | } | |
972 | ||
07f1e596 AG |
973 | ofs->fs[ofs->numfs].sb = sb; |
974 | ofs->fs[ofs->numfs].pseudo_dev = dev; | |
975 | ofs->fs[ofs->numfs].bad_uuid = bad_uuid; | |
5148626b | 976 | |
07f1e596 | 977 | return ofs->numfs++; |
5148626b AG |
978 | } |
979 | ||
37ebf056 AG |
980 | /* |
981 | * The fsid after the last lower fsid is used for the data layers. | |
982 | * It is a "null fs" with a null sb, null uuid, and no pseudo dev. | |
983 | */ | |
984 | static int ovl_get_data_fsid(struct ovl_fs *ofs) | |
985 | { | |
986 | return ofs->numfs; | |
987 | } | |
988 | ||
989 | ||
94375f9d | 990 | static int ovl_get_layers(struct super_block *sb, struct ovl_fs *ofs, |
b36a5780 | 991 | struct ovl_fs_context *ctx, struct ovl_layer *layers) |
520d7c86 MS |
992 | { |
993 | int err; | |
994 | unsigned int i; | |
b36a5780 | 995 | size_t nr_merged_lower; |
520d7c86 | 996 | |
b36a5780 | 997 | ofs->fs = kcalloc(ctx->nr + 2, sizeof(struct ovl_sb), GFP_KERNEL); |
07f1e596 | 998 | if (ofs->fs == NULL) |
9e88f905 | 999 | return -ENOMEM; |
5148626b | 1000 | |
37ebf056 AG |
1001 | /* |
1002 | * idx/fsid 0 are reserved for upper fs even with lower only overlay | |
1003 | * and the last fsid is reserved for "null fs" of the data layers. | |
1004 | */ | |
07f1e596 AG |
1005 | ofs->numfs++; |
1006 | ||
07f1e596 | 1007 | /* |
b7bf9908 AG |
1008 | * All lower layers that share the same fs as upper layer, use the same |
1009 | * pseudo_dev as upper layer. Allocate fs[0].pseudo_dev even for lower | |
1010 | * only overlay to simplify ovl_fs_free(). | |
1b81dddd | 1011 | * is_lower will be set if upper fs is shared with a lower layer. |
07f1e596 | 1012 | */ |
b7bf9908 AG |
1013 | err = get_anon_bdev(&ofs->fs[0].pseudo_dev); |
1014 | if (err) { | |
1015 | pr_err("failed to get anonymous bdev for upper fs\n"); | |
9e88f905 | 1016 | return err; |
b7bf9908 AG |
1017 | } |
1018 | ||
08f4c7c8 MS |
1019 | if (ovl_upper_mnt(ofs)) { |
1020 | ofs->fs[0].sb = ovl_upper_mnt(ofs)->mnt_sb; | |
1b81dddd | 1021 | ofs->fs[0].is_lower = false; |
07f1e596 AG |
1022 | } |
1023 | ||
b36a5780 CB |
1024 | nr_merged_lower = ctx->nr - ctx->nr_data; |
1025 | for (i = 0; i < ctx->nr; i++) { | |
1026 | struct ovl_fs_context_layer *l = &ctx->lower[i]; | |
520d7c86 | 1027 | struct vfsmount *mnt; |
146d62e5 | 1028 | struct inode *trap; |
5148626b | 1029 | int fsid; |
520d7c86 | 1030 | |
b36a5780 CB |
1031 | if (i < nr_merged_lower) |
1032 | fsid = ovl_get_fsid(ofs, &l->path); | |
37ebf056 AG |
1033 | else |
1034 | fsid = ovl_get_data_fsid(ofs); | |
9e88f905 AG |
1035 | if (fsid < 0) |
1036 | return fsid; | |
520d7c86 | 1037 | |
24f14009 | 1038 | /* |
1039 | * Check if lower root conflicts with this overlay layers before | |
1040 | * checking if it is in-use as upperdir/workdir of "another" | |
1041 | * mount, because we do not bother to check in ovl_is_inuse() if | |
1042 | * the upperdir/workdir is in fact in-use by our | |
1043 | * upperdir/workdir. | |
1044 | */ | |
b36a5780 | 1045 | err = ovl_setup_trap(sb, l->path.dentry, &trap, "lowerdir"); |
146d62e5 | 1046 | if (err) |
9e88f905 | 1047 | return err; |
146d62e5 | 1048 | |
b36a5780 | 1049 | if (ovl_is_inuse(l->path.dentry)) { |
0be0bfd2 | 1050 | err = ovl_report_in_use(ofs, "lowerdir"); |
24f14009 | 1051 | if (err) { |
1052 | iput(trap); | |
9e88f905 | 1053 | return err; |
24f14009 | 1054 | } |
0be0bfd2 AG |
1055 | } |
1056 | ||
b36a5780 | 1057 | mnt = clone_private_mount(&l->path); |
520d7c86 MS |
1058 | err = PTR_ERR(mnt); |
1059 | if (IS_ERR(mnt)) { | |
1bd0a3ae | 1060 | pr_err("failed to clone lowerpath\n"); |
146d62e5 | 1061 | iput(trap); |
9e88f905 | 1062 | return err; |
520d7c86 | 1063 | } |
5148626b | 1064 | |
520d7c86 MS |
1065 | /* |
1066 | * Make lower layers R/O. That way fchmod/fchown on lower file | |
1067 | * will fail instead of modifying lower fs. | |
1068 | */ | |
1069 | mnt->mnt_flags |= MNT_READONLY | MNT_NOATIME; | |
1070 | ||
13464165 MS |
1071 | layers[ofs->numlayer].trap = trap; |
1072 | layers[ofs->numlayer].mnt = mnt; | |
1073 | layers[ofs->numlayer].idx = ofs->numlayer; | |
1074 | layers[ofs->numlayer].fsid = fsid; | |
1075 | layers[ofs->numlayer].fs = &ofs->fs[fsid]; | |
a535116d AG |
1076 | /* Store for printing lowerdir=... in ovl_show_options() */ |
1077 | ofs->config.lowerdirs[ofs->numlayer] = l->name; | |
b36a5780 | 1078 | l->name = NULL; |
94375f9d | 1079 | ofs->numlayer++; |
1b81dddd | 1080 | ofs->fs[fsid].is_lower = true; |
520d7c86 | 1081 | } |
e487d889 | 1082 | |
795939a9 AG |
1083 | /* |
1084 | * When all layers on same fs, overlay can use real inode numbers. | |
926e94d7 AG |
1085 | * With mount option "xino=<on|auto>", mounter declares that there are |
1086 | * enough free high bits in underlying fs to hold the unique fsid. | |
795939a9 AG |
1087 | * If overlayfs does encounter underlying inodes using the high xino |
1088 | * bits reserved for fsid, it emits a warning and uses the original | |
dfe51d47 AG |
1089 | * inode number or a non persistent inode number allocated from a |
1090 | * dedicated range. | |
795939a9 | 1091 | */ |
08f4c7c8 | 1092 | if (ofs->numfs - !ovl_upper_mnt(ofs) == 1) { |
0f831ec8 AG |
1093 | if (ofs->config.xino == OVL_XINO_ON) |
1094 | pr_info("\"xino=on\" is useless with all layers on same fs, ignore.\n"); | |
1095 | ofs->xino_mode = 0; | |
53afcd31 AG |
1096 | } else if (ofs->config.xino == OVL_XINO_OFF) { |
1097 | ofs->xino_mode = -1; | |
926e94d7 | 1098 | } else if (ofs->xino_mode < 0) { |
795939a9 | 1099 | /* |
07f1e596 | 1100 | * This is a roundup of number of bits needed for encoding |
dfe51d47 AG |
1101 | * fsid, where fsid 0 is reserved for upper fs (even with |
1102 | * lower only overlay) +1 extra bit is reserved for the non | |
1103 | * persistent inode number range that is used for resolving | |
1104 | * xino lower bits overflow. | |
795939a9 | 1105 | */ |
dfe51d47 AG |
1106 | BUILD_BUG_ON(ilog2(OVL_MAX_STACK) > 30); |
1107 | ofs->xino_mode = ilog2(ofs->numfs - 1) + 2; | |
795939a9 AG |
1108 | } |
1109 | ||
0f831ec8 | 1110 | if (ofs->xino_mode > 0) { |
1bd0a3ae | 1111 | pr_info("\"xino\" feature enabled using %d upper inode bits.\n", |
0f831ec8 | 1112 | ofs->xino_mode); |
795939a9 | 1113 | } |
e487d889 | 1114 | |
9e88f905 | 1115 | return 0; |
520d7c86 MS |
1116 | } |
1117 | ||
4155c10a | 1118 | static struct ovl_entry *ovl_get_lowerstack(struct super_block *sb, |
b36a5780 CB |
1119 | struct ovl_fs_context *ctx, |
1120 | struct ovl_fs *ofs, | |
1121 | struct ovl_layer *layers) | |
53dbb0b4 MS |
1122 | { |
1123 | int err; | |
b8e42a65 | 1124 | unsigned int i; |
b36a5780 | 1125 | size_t nr_merged_lower; |
4155c10a | 1126 | struct ovl_entry *oe; |
b36a5780 CB |
1127 | struct ovl_path *lowerstack; |
1128 | ||
1129 | struct ovl_fs_context_layer *l; | |
53dbb0b4 | 1130 | |
b36a5780 | 1131 | if (!ofs->config.upperdir && ctx->nr == 1) { |
1bd0a3ae | 1132 | pr_err("at least 2 lowerdir are needed while upperdir nonexistent\n"); |
b8e42a65 | 1133 | return ERR_PTR(-EINVAL); |
53dbb0b4 MS |
1134 | } |
1135 | ||
b36a5780 CB |
1136 | err = -EINVAL; |
1137 | for (i = 0; i < ctx->nr; i++) { | |
1138 | l = &ctx->lower[i]; | |
53dbb0b4 | 1139 | |
b36a5780 | 1140 | err = ovl_lower_dir(l->name, &l->path, ofs, &sb->s_stack_depth); |
53dbb0b4 | 1141 | if (err) |
b36a5780 | 1142 | return ERR_PTR(err); |
53dbb0b4 MS |
1143 | } |
1144 | ||
1145 | err = -EINVAL; | |
1146 | sb->s_stack_depth++; | |
1147 | if (sb->s_stack_depth > FILESYSTEM_MAX_STACK_DEPTH) { | |
1bd0a3ae | 1148 | pr_err("maximum fs stacking depth exceeded\n"); |
b36a5780 | 1149 | return ERR_PTR(err); |
53dbb0b4 MS |
1150 | } |
1151 | ||
b36a5780 | 1152 | err = ovl_get_layers(sb, ofs, ctx, layers); |
4155c10a | 1153 | if (err) |
b36a5780 | 1154 | return ERR_PTR(err); |
4155c10a MS |
1155 | |
1156 | err = -ENOMEM; | |
37ebf056 | 1157 | /* Data-only layers are not merged in root directory */ |
b36a5780 CB |
1158 | nr_merged_lower = ctx->nr - ctx->nr_data; |
1159 | oe = ovl_alloc_entry(nr_merged_lower); | |
4155c10a | 1160 | if (!oe) |
b36a5780 | 1161 | return ERR_PTR(err); |
4155c10a | 1162 | |
5522c9c7 | 1163 | lowerstack = ovl_lowerstack(oe); |
b36a5780 CB |
1164 | for (i = 0; i < nr_merged_lower; i++) { |
1165 | l = &ctx->lower[i]; | |
1166 | lowerstack[i].dentry = dget(l->path.dentry); | |
1167 | lowerstack[i].layer = &ofs->layers[i + 1]; | |
4155c10a | 1168 | } |
b36a5780 | 1169 | ofs->numdatalayer = ctx->nr_data; |
4155c10a MS |
1170 | |
1171 | return oe; | |
53dbb0b4 MS |
1172 | } |
1173 | ||
146d62e5 AG |
1174 | /* |
1175 | * Check if this layer root is a descendant of: | |
1176 | * - another layer of this overlayfs instance | |
1177 | * - upper/work dir of any overlayfs instance | |
146d62e5 | 1178 | */ |
0be0bfd2 | 1179 | static int ovl_check_layer(struct super_block *sb, struct ovl_fs *ofs, |
708fa015 MS |
1180 | struct dentry *dentry, const char *name, |
1181 | bool is_lower) | |
146d62e5 | 1182 | { |
9179c21d | 1183 | struct dentry *next = dentry, *parent; |
146d62e5 AG |
1184 | int err = 0; |
1185 | ||
9179c21d | 1186 | if (!dentry) |
146d62e5 AG |
1187 | return 0; |
1188 | ||
9179c21d MS |
1189 | parent = dget_parent(next); |
1190 | ||
1191 | /* Walk back ancestors to root (inclusive) looking for traps */ | |
1192 | while (!err && parent != next) { | |
708fa015 | 1193 | if (is_lower && ovl_lookup_trap_inode(sb, parent)) { |
146d62e5 | 1194 | err = -ELOOP; |
1bd0a3ae | 1195 | pr_err("overlapping %s path\n", name); |
0be0bfd2 AG |
1196 | } else if (ovl_is_inuse(parent)) { |
1197 | err = ovl_report_in_use(ofs, name); | |
146d62e5 | 1198 | } |
146d62e5 | 1199 | next = parent; |
9179c21d MS |
1200 | parent = dget_parent(next); |
1201 | dput(next); | |
146d62e5 AG |
1202 | } |
1203 | ||
9179c21d | 1204 | dput(parent); |
146d62e5 AG |
1205 | |
1206 | return err; | |
1207 | } | |
1208 | ||
1209 | /* | |
1210 | * Check if any of the layers or work dirs overlap. | |
1211 | */ | |
1212 | static int ovl_check_overlapping_layers(struct super_block *sb, | |
1213 | struct ovl_fs *ofs) | |
1214 | { | |
1215 | int i, err; | |
1216 | ||
08f4c7c8 MS |
1217 | if (ovl_upper_mnt(ofs)) { |
1218 | err = ovl_check_layer(sb, ofs, ovl_upper_mnt(ofs)->mnt_root, | |
708fa015 | 1219 | "upperdir", false); |
146d62e5 AG |
1220 | if (err) |
1221 | return err; | |
1222 | ||
1223 | /* | |
1224 | * Checking workbasedir avoids hitting ovl_is_inuse(parent) of | |
1225 | * this instance and covers overlapping work and index dirs, | |
1226 | * unless work or index dir have been moved since created inside | |
1227 | * workbasedir. In that case, we already have their traps in | |
1228 | * inode cache and we will catch that case on lookup. | |
1229 | */ | |
708fa015 MS |
1230 | err = ovl_check_layer(sb, ofs, ofs->workbasedir, "workdir", |
1231 | false); | |
146d62e5 AG |
1232 | if (err) |
1233 | return err; | |
1234 | } | |
1235 | ||
94375f9d | 1236 | for (i = 1; i < ofs->numlayer; i++) { |
0be0bfd2 | 1237 | err = ovl_check_layer(sb, ofs, |
94375f9d | 1238 | ofs->layers[i].mnt->mnt_root, |
708fa015 | 1239 | "lowerdir", true); |
146d62e5 AG |
1240 | if (err) |
1241 | return err; | |
1242 | } | |
1243 | ||
1244 | return 0; | |
1245 | } | |
1246 | ||
2effc5c2 AG |
1247 | static struct dentry *ovl_get_root(struct super_block *sb, |
1248 | struct dentry *upperdentry, | |
1249 | struct ovl_entry *oe) | |
1250 | { | |
1251 | struct dentry *root; | |
420332b9 | 1252 | struct ovl_fs *ofs = OVL_FS(sb); |
5522c9c7 | 1253 | struct ovl_path *lowerpath = ovl_lowerstack(oe); |
62c832ed AG |
1254 | unsigned long ino = d_inode(lowerpath->dentry)->i_ino; |
1255 | int fsid = lowerpath->layer->fsid; | |
1256 | struct ovl_inode_params oip = { | |
1257 | .upperdentry = upperdentry, | |
0af950f5 | 1258 | .oe = oe, |
62c832ed | 1259 | }; |
2effc5c2 AG |
1260 | |
1261 | root = d_make_root(ovl_new_inode(sb, S_IFDIR, 0)); | |
1262 | if (!root) | |
1263 | return NULL; | |
1264 | ||
2effc5c2 | 1265 | if (upperdentry) { |
62c832ed AG |
1266 | /* Root inode uses upper st_ino/i_ino */ |
1267 | ino = d_inode(upperdentry)->i_ino; | |
1268 | fsid = 0; | |
2effc5c2 | 1269 | ovl_dentry_set_upper_alias(root); |
610afc0b | 1270 | if (ovl_is_impuredir(sb, upperdentry)) |
2effc5c2 AG |
1271 | ovl_set_flag(OVL_IMPURE, d_inode(root)); |
1272 | } | |
1273 | ||
420332b9 AG |
1274 | /* Look for xwhiteouts marker except in the lowermost layer */ |
1275 | for (int i = 0; i < ovl_numlower(oe) - 1; i++, lowerpath++) { | |
1276 | struct path path = { | |
1277 | .mnt = lowerpath->layer->mnt, | |
1278 | .dentry = lowerpath->dentry, | |
1279 | }; | |
1280 | ||
1281 | /* overlay.opaque=x means xwhiteouts directory */ | |
1282 | if (ovl_get_opaquedir_val(ofs, &path) == 'x') { | |
1283 | ovl_layer_set_xwhiteouts(ofs, lowerpath->layer); | |
1284 | ovl_dentry_set_xwhiteouts(root); | |
1285 | } | |
1286 | } | |
1287 | ||
2effc5c2 AG |
1288 | /* Root is always merge -> can have whiteouts */ |
1289 | ovl_set_flag(OVL_WHITEOUTS, d_inode(root)); | |
1290 | ovl_dentry_set_flag(OVL_E_CONNECTED, root); | |
1291 | ovl_set_upperdata(d_inode(root)); | |
62c832ed | 1292 | ovl_inode_init(d_inode(root), &oip, ino, fsid); |
0af950f5 | 1293 | ovl_dentry_init_flags(root, upperdentry, oe, DCACHE_OP_WEAK_REVALIDATE); |
367d002d AG |
1294 | /* root keeps a reference of upperdentry */ |
1295 | dget(upperdentry); | |
2effc5c2 AG |
1296 | |
1297 | return root; | |
1298 | } | |
1299 | ||
7fb7998b | 1300 | int ovl_fill_super(struct super_block *sb, struct fs_context *fc) |
e9be9d5e | 1301 | { |
1784fbc2 CB |
1302 | struct ovl_fs *ofs = sb->s_fs_info; |
1303 | struct ovl_fs_context *ctx = fc->fs_private; | |
e9be9d5e | 1304 | struct dentry *root_dentry; |
4155c10a | 1305 | struct ovl_entry *oe; |
b8e42a65 | 1306 | struct ovl_layer *layers; |
51f8f3c4 | 1307 | struct cred *cred; |
e9be9d5e MS |
1308 | int err; |
1309 | ||
9efb069d | 1310 | err = -EIO; |
1784fbc2 CB |
1311 | if (WARN_ON(fc->user_ns != current_user_ns())) |
1312 | goto out_err; | |
9efb069d | 1313 | |
f4288844 MS |
1314 | sb->s_d_op = &ovl_dentry_operations; |
1315 | ||
d7b49b10 | 1316 | err = -ENOMEM; |
ad204488 | 1317 | ofs->creator_cred = cred = prepare_creds(); |
c6fe6254 MS |
1318 | if (!cred) |
1319 | goto out_err; | |
1320 | ||
1784fbc2 | 1321 | err = ovl_fs_params_verify(ctx, &ofs->config); |
f45827e8 | 1322 | if (err) |
a9075cdb | 1323 | goto out_err; |
f45827e8 | 1324 | |
e9be9d5e | 1325 | err = -EINVAL; |
b36a5780 | 1326 | if (ctx->nr == 0) { |
1784fbc2 | 1327 | if (!(fc->sb_flags & SB_SILENT)) |
1bd0a3ae | 1328 | pr_err("missing 'lowerdir'\n"); |
a9075cdb | 1329 | goto out_err; |
e9be9d5e MS |
1330 | } |
1331 | ||
b8e42a65 | 1332 | err = -ENOMEM; |
b36a5780 | 1333 | layers = kcalloc(ctx->nr + 1, sizeof(struct ovl_layer), GFP_KERNEL); |
b8e42a65 MS |
1334 | if (!layers) |
1335 | goto out_err; | |
1336 | ||
a535116d AG |
1337 | ofs->config.lowerdirs = kcalloc(ctx->nr + 1, sizeof(char *), GFP_KERNEL); |
1338 | if (!ofs->config.lowerdirs) { | |
1339 | kfree(layers); | |
1340 | goto out_err; | |
1341 | } | |
b8e42a65 | 1342 | ofs->layers = layers; |
a535116d AG |
1343 | /* |
1344 | * Layer 0 is reserved for upper even if there's no upper. | |
0cea4c09 AG |
1345 | * config.lowerdirs[0] is used for storing the user provided colon |
1346 | * separated lowerdir string. | |
a535116d | 1347 | */ |
0cea4c09 AG |
1348 | ofs->config.lowerdirs[0] = ctx->lowerdir_all; |
1349 | ctx->lowerdir_all = NULL; | |
b8e42a65 MS |
1350 | ofs->numlayer = 1; |
1351 | ||
53a08cb9 | 1352 | sb->s_stack_depth = 0; |
cf9a6784 | 1353 | sb->s_maxbytes = MAX_LFS_FILESIZE; |
4d314f78 | 1354 | atomic_long_set(&ofs->last_ino, 1); |
4f119628 | 1355 | /* Assume underlying fs uses 32bit inodes unless proven otherwise */ |
53afcd31 | 1356 | if (ofs->config.xino != OVL_XINO_OFF) { |
0f831ec8 | 1357 | ofs->xino_mode = BITS_PER_LONG - 32; |
53afcd31 AG |
1358 | if (!ofs->xino_mode) { |
1359 | pr_warn("xino not supported on 32bit kernel, falling back to xino=off.\n"); | |
1360 | ofs->config.xino = OVL_XINO_OFF; | |
1361 | } | |
1362 | } | |
795939a9 | 1363 | |
146d62e5 AG |
1364 | /* alloc/destroy_inode needed for setting up traps in inode cache */ |
1365 | sb->s_op = &ovl_super_operations; | |
1366 | ||
ad204488 | 1367 | if (ofs->config.upperdir) { |
335d3fc5 SD |
1368 | struct super_block *upper_sb; |
1369 | ||
d7b49b10 | 1370 | err = -EINVAL; |
ad204488 | 1371 | if (!ofs->config.workdir) { |
1bd0a3ae | 1372 | pr_err("missing 'workdir'\n"); |
a9075cdb | 1373 | goto out_err; |
53a08cb9 | 1374 | } |
e9be9d5e | 1375 | |
b36a5780 | 1376 | err = ovl_get_upper(sb, ofs, &layers[0], &ctx->upper); |
53a08cb9 | 1377 | if (err) |
a9075cdb | 1378 | goto out_err; |
2cac0c00 | 1379 | |
335d3fc5 SD |
1380 | upper_sb = ovl_upper_mnt(ofs)->mnt_sb; |
1381 | if (!ovl_should_sync(ofs)) { | |
1382 | ofs->errseq = errseq_sample(&upper_sb->s_wb_err); | |
1383 | if (errseq_check(&upper_sb->s_wb_err, ofs->errseq)) { | |
1384 | err = -EIO; | |
1385 | pr_err("Cannot mount volatile when upperdir has an unseen error. Sync upperdir fs to clear state.\n"); | |
1386 | goto out_err; | |
1387 | } | |
1388 | } | |
1389 | ||
b36a5780 | 1390 | err = ovl_get_workdir(sb, ofs, &ctx->upper, &ctx->work); |
8ed61dc3 | 1391 | if (err) |
a9075cdb | 1392 | goto out_err; |
c6fe6254 | 1393 | |
ad204488 | 1394 | if (!ofs->workdir) |
1751e8a6 | 1395 | sb->s_flags |= SB_RDONLY; |
6e88256e | 1396 | |
335d3fc5 SD |
1397 | sb->s_stack_depth = upper_sb->s_stack_depth; |
1398 | sb->s_time_gran = upper_sb->s_time_gran; | |
e9be9d5e | 1399 | } |
b36a5780 | 1400 | oe = ovl_get_lowerstack(sb, ctx, ofs, layers); |
4155c10a MS |
1401 | err = PTR_ERR(oe); |
1402 | if (IS_ERR(oe)) | |
a9075cdb | 1403 | goto out_err; |
e9be9d5e | 1404 | |
71cbad7e | 1405 | /* If the upper fs is nonexistent, we mark overlayfs r/o too */ |
08f4c7c8 | 1406 | if (!ovl_upper_mnt(ofs)) |
1751e8a6 | 1407 | sb->s_flags |= SB_RDONLY; |
e9be9d5e | 1408 | |
b0504bfe AG |
1409 | if (!ovl_origin_uuid(ofs) && ofs->numfs > 1) { |
1410 | pr_warn("The uuid=off requires a single fs for lower and upper, falling back to uuid=null.\n"); | |
1411 | ofs->config.uuid = OVL_UUID_NULL; | |
d9544c1b AG |
1412 | } else if (ovl_has_fsid(ofs) && ovl_upper_mnt(ofs)) { |
1413 | /* Use per instance persistent uuid/fsid */ | |
1414 | ovl_init_uuid_xattr(sb, ofs, &ctx->upper); | |
5830fb6b PT |
1415 | } |
1416 | ||
470c1563 | 1417 | if (!ovl_force_readonly(ofs) && ofs->config.index) { |
b36a5780 | 1418 | err = ovl_get_indexdir(sb, ofs, oe, &ctx->upper); |
54fb347e | 1419 | if (err) |
4155c10a | 1420 | goto out_free_oe; |
6e88256e | 1421 | |
972d0093 | 1422 | /* Force r/o mount with no index dir */ |
02d70090 | 1423 | if (!ofs->workdir) |
1751e8a6 | 1424 | sb->s_flags |= SB_RDONLY; |
02bcd157 AG |
1425 | } |
1426 | ||
146d62e5 AG |
1427 | err = ovl_check_overlapping_layers(sb, ofs); |
1428 | if (err) | |
1429 | goto out_free_oe; | |
1430 | ||
972d0093 | 1431 | /* Show index=off in /proc/mounts for forced r/o mount */ |
02d70090 | 1432 | if (!ofs->workdir) { |
ad204488 | 1433 | ofs->config.index = false; |
08f4c7c8 | 1434 | if (ovl_upper_mnt(ofs) && ofs->config.nfs_export) { |
1bd0a3ae | 1435 | pr_warn("NFS export requires an index dir, falling back to nfs_export=off.\n"); |
f168f109 AG |
1436 | ofs->config.nfs_export = false; |
1437 | } | |
1438 | } | |
02bcd157 | 1439 | |
d5791044 | 1440 | if (ofs->config.metacopy && ofs->config.nfs_export) { |
1bd0a3ae | 1441 | pr_warn("NFS export is not supported with metadata only copy up, falling back to nfs_export=off.\n"); |
d5791044 VG |
1442 | ofs->config.nfs_export = false; |
1443 | } | |
1444 | ||
16aac5ad AG |
1445 | /* |
1446 | * Support encoding decodable file handles with nfs_export=on | |
1447 | * and encoding non-decodable file handles with nfs_export=off | |
1448 | * if all layers support file handles. | |
1449 | */ | |
8383f174 AG |
1450 | if (ofs->config.nfs_export) |
1451 | sb->s_export_op = &ovl_export_operations; | |
16aac5ad AG |
1452 | else if (!ofs->nofh) |
1453 | sb->s_export_op = &ovl_export_fid_operations; | |
8383f174 | 1454 | |
51f8f3c4 KK |
1455 | /* Never override disk quota limits or use reserved space */ |
1456 | cap_lower(cred->cap_effective, CAP_SYS_RESOURCE); | |
1457 | ||
655042cc | 1458 | sb->s_magic = OVERLAYFS_SUPER_MAGIC; |
420a62dd | 1459 | sb->s_xattr = ovl_xattr_handlers(ofs); |
ad204488 | 1460 | sb->s_fs_info = ofs; |
2bc5e5e8 | 1461 | #ifdef CONFIG_FS_POSIX_ACL |
7c4d37c2 | 1462 | sb->s_flags |= SB_POSIXACL; |
2bc5e5e8 | 1463 | #endif |
b836c4d2 | 1464 | sb->s_iflags |= SB_I_SKIP_SYNC; |
2bc5e5e8 CB |
1465 | /* |
1466 | * Ensure that umask handling is done by the filesystems used | |
1467 | * for the the upper layer instead of overlayfs as that would | |
1468 | * lead to unexpected results. | |
1469 | */ | |
1470 | sb->s_iflags |= SB_I_NOUMASK; | |
c00f94b3 | 1471 | sb->s_iflags |= SB_I_EVM_UNSUPPORTED; |
655042cc | 1472 | |
c6fe6254 | 1473 | err = -ENOMEM; |
b36a5780 | 1474 | root_dentry = ovl_get_root(sb, ctx->upper.dentry, oe); |
e9be9d5e | 1475 | if (!root_dentry) |
4155c10a | 1476 | goto out_free_oe; |
e9be9d5e | 1477 | |
e9be9d5e | 1478 | sb->s_root = root_dentry; |
e9be9d5e MS |
1479 | |
1480 | return 0; | |
1481 | ||
4155c10a | 1482 | out_free_oe: |
163db0da | 1483 | ovl_free_entry(oe); |
4155c10a | 1484 | out_err: |
ad204488 | 1485 | ovl_free_fs(ofs); |
1784fbc2 | 1486 | sb->s_fs_info = NULL; |
e9be9d5e MS |
1487 | return err; |
1488 | } | |
1489 | ||
f01d0889 | 1490 | struct file_system_type ovl_fs_type = { |
1784fbc2 CB |
1491 | .owner = THIS_MODULE, |
1492 | .name = "overlay", | |
1493 | .init_fs_context = ovl_init_fs_context, | |
1494 | .parameters = ovl_parameter_spec, | |
1495 | .fs_flags = FS_USERNS_MOUNT, | |
1496 | .kill_sb = kill_anon_super, | |
e9be9d5e | 1497 | }; |
ef94b186 | 1498 | MODULE_ALIAS_FS("overlay"); |
e9be9d5e | 1499 | |
13cf199d AG |
1500 | static void ovl_inode_init_once(void *foo) |
1501 | { | |
1502 | struct ovl_inode *oi = foo; | |
1503 | ||
1504 | inode_init_once(&oi->vfs_inode); | |
1505 | } | |
1506 | ||
e9be9d5e MS |
1507 | static int __init ovl_init(void) |
1508 | { | |
13cf199d AG |
1509 | int err; |
1510 | ||
1511 | ovl_inode_cachep = kmem_cache_create("ovl_inode", | |
1512 | sizeof(struct ovl_inode), 0, | |
1513 | (SLAB_RECLAIM_ACCOUNT| | |
1514 | SLAB_MEM_SPREAD|SLAB_ACCOUNT), | |
1515 | ovl_inode_init_once); | |
1516 | if (ovl_inode_cachep == NULL) | |
1517 | return -ENOMEM; | |
1518 | ||
a6293b3e AG |
1519 | err = register_filesystem(&ovl_fs_type); |
1520 | if (!err) | |
1521 | return 0; | |
2406a307 | 1522 | |
2406a307 | 1523 | kmem_cache_destroy(ovl_inode_cachep); |
13cf199d AG |
1524 | |
1525 | return err; | |
e9be9d5e MS |
1526 | } |
1527 | ||
1528 | static void __exit ovl_exit(void) | |
1529 | { | |
1530 | unregister_filesystem(&ovl_fs_type); | |
13cf199d AG |
1531 | |
1532 | /* | |
1533 | * Make sure all delayed rcu free inodes are flushed before we | |
1534 | * destroy cache. | |
1535 | */ | |
1536 | rcu_barrier(); | |
1537 | kmem_cache_destroy(ovl_inode_cachep); | |
e9be9d5e MS |
1538 | } |
1539 | ||
1540 | module_init(ovl_init); | |
1541 | module_exit(ovl_exit); |