Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * linux/fs/namei.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | */ | |
7 | ||
8 | /* | |
9 | * Some corrections by tytso. | |
10 | */ | |
11 | ||
12 | /* [Feb 1997 T. Schoebel-Theuer] Complete rewrite of the pathname | |
13 | * lookup logic. | |
14 | */ | |
15 | /* [Feb-Apr 2000, AV] Rewrite to the new namespace architecture. | |
16 | */ | |
17 | ||
18 | #include <linux/init.h> | |
630d9c47 | 19 | #include <linux/export.h> |
44696908 | 20 | #include <linux/kernel.h> |
1da177e4 LT |
21 | #include <linux/slab.h> |
22 | #include <linux/fs.h> | |
23 | #include <linux/namei.h> | |
1da177e4 | 24 | #include <linux/pagemap.h> |
0eeca283 | 25 | #include <linux/fsnotify.h> |
1da177e4 LT |
26 | #include <linux/personality.h> |
27 | #include <linux/security.h> | |
6146f0d5 | 28 | #include <linux/ima.h> |
1da177e4 LT |
29 | #include <linux/syscalls.h> |
30 | #include <linux/mount.h> | |
31 | #include <linux/audit.h> | |
16f7e0fe | 32 | #include <linux/capability.h> |
834f2a4a | 33 | #include <linux/file.h> |
5590ff0d | 34 | #include <linux/fcntl.h> |
08ce5f16 | 35 | #include <linux/device_cgroup.h> |
5ad4e53b | 36 | #include <linux/fs_struct.h> |
e77819e5 | 37 | #include <linux/posix_acl.h> |
99d263d4 | 38 | #include <linux/hash.h> |
2a18da7a | 39 | #include <linux/bitops.h> |
aeaa4a79 | 40 | #include <linux/init_task.h> |
7c0f6ba6 | 41 | #include <linux/uaccess.h> |
1da177e4 | 42 | |
e81e3f4d | 43 | #include "internal.h" |
c7105365 | 44 | #include "mount.h" |
e81e3f4d | 45 | |
1da177e4 LT |
46 | /* [Feb-1997 T. Schoebel-Theuer] |
47 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
48 | * were necessary because of omirr. The reason is that omirr needs | |
49 | * to know the _real_ pathname, not the user-supplied one, in case | |
50 | * of symlinks (and also when transname replacements occur). | |
51 | * | |
52 | * The new code replaces the old recursive symlink resolution with | |
53 | * an iterative one (in case of non-nested symlink chains). It does | |
54 | * this with calls to <fs>_follow_link(). | |
55 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
56 | * replaced with a single function lookup_dentry() that can handle all | |
57 | * the special cases of the former code. | |
58 | * | |
59 | * With the new dcache, the pathname is stored at each inode, at least as | |
60 | * long as the refcount of the inode is positive. As a side effect, the | |
61 | * size of the dcache depends on the inode cache and thus is dynamic. | |
62 | * | |
63 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
64 | * resolution to correspond with current state of the code. | |
65 | * | |
66 | * Note that the symlink resolution is not *completely* iterative. | |
67 | * There is still a significant amount of tail- and mid- recursion in | |
68 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
69 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
70 | * may return different results than <fs>_follow_link(). Many virtual | |
71 | * filesystems (including /proc) exhibit this behavior. | |
72 | */ | |
73 | ||
74 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
75 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
76 | * and the name already exists in form of a symlink, try to create the new | |
77 | * name indicated by the symlink. The old code always complained that the | |
78 | * name already exists, due to not following the symlink even if its target | |
79 | * is nonexistent. The new semantics affects also mknod() and link() when | |
25985edc | 80 | * the name is a symlink pointing to a non-existent name. |
1da177e4 LT |
81 | * |
82 | * I don't know which semantics is the right one, since I have no access | |
83 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
84 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
85 | * "old" one. Personally, I think the new semantics is much more logical. | |
86 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
87 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
88 | * and in the old Linux semantics. | |
89 | */ | |
90 | ||
91 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
92 | * semantics. See the comments in "open_namei" and "do_link" below. | |
93 | * | |
94 | * [10-Sep-98 Alan Modra] Another symlink change. | |
95 | */ | |
96 | ||
97 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
98 | * inside the path - always follow. | |
99 | * in the last component in creation/removal/renaming - never follow. | |
100 | * if LOOKUP_FOLLOW passed - follow. | |
101 | * if the pathname has trailing slashes - follow. | |
102 | * otherwise - don't follow. | |
103 | * (applied in that order). | |
104 | * | |
105 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
106 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
107 | * During the 2.4 we need to fix the userland stuff depending on it - | |
108 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
109 | * XEmacs seems to be relying on it... | |
110 | */ | |
111 | /* | |
112 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 113 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
114 | * any extra contention... |
115 | */ | |
116 | ||
117 | /* In order to reduce some races, while at the same time doing additional | |
118 | * checking and hopefully speeding things up, we copy filenames to the | |
119 | * kernel data space before using them.. | |
120 | * | |
121 | * POSIX.1 2.4: an empty pathname is invalid (ENOENT). | |
122 | * PATH_MAX includes the nul terminator --RR. | |
123 | */ | |
91a27b2a | 124 | |
fd2f7cb5 | 125 | #define EMBEDDED_NAME_MAX (PATH_MAX - offsetof(struct filename, iname)) |
7950e385 | 126 | |
51f39a1f | 127 | struct filename * |
91a27b2a JL |
128 | getname_flags(const char __user *filename, int flags, int *empty) |
129 | { | |
94b5d262 | 130 | struct filename *result; |
7950e385 | 131 | char *kname; |
94b5d262 | 132 | int len; |
4043cde8 | 133 | |
7ac86265 JL |
134 | result = audit_reusename(filename); |
135 | if (result) | |
136 | return result; | |
137 | ||
7950e385 | 138 | result = __getname(); |
3f9f0aa6 | 139 | if (unlikely(!result)) |
4043cde8 EP |
140 | return ERR_PTR(-ENOMEM); |
141 | ||
7950e385 JL |
142 | /* |
143 | * First, try to embed the struct filename inside the names_cache | |
144 | * allocation | |
145 | */ | |
fd2f7cb5 | 146 | kname = (char *)result->iname; |
91a27b2a | 147 | result->name = kname; |
7950e385 | 148 | |
94b5d262 | 149 | len = strncpy_from_user(kname, filename, EMBEDDED_NAME_MAX); |
91a27b2a | 150 | if (unlikely(len < 0)) { |
94b5d262 AV |
151 | __putname(result); |
152 | return ERR_PTR(len); | |
91a27b2a | 153 | } |
3f9f0aa6 | 154 | |
7950e385 JL |
155 | /* |
156 | * Uh-oh. We have a name that's approaching PATH_MAX. Allocate a | |
157 | * separate struct filename so we can dedicate the entire | |
158 | * names_cache allocation for the pathname, and re-do the copy from | |
159 | * userland. | |
160 | */ | |
94b5d262 | 161 | if (unlikely(len == EMBEDDED_NAME_MAX)) { |
fd2f7cb5 | 162 | const size_t size = offsetof(struct filename, iname[1]); |
7950e385 JL |
163 | kname = (char *)result; |
164 | ||
fd2f7cb5 AV |
165 | /* |
166 | * size is chosen that way we to guarantee that | |
167 | * result->iname[0] is within the same object and that | |
168 | * kname can't be equal to result->iname, no matter what. | |
169 | */ | |
170 | result = kzalloc(size, GFP_KERNEL); | |
94b5d262 AV |
171 | if (unlikely(!result)) { |
172 | __putname(kname); | |
173 | return ERR_PTR(-ENOMEM); | |
7950e385 JL |
174 | } |
175 | result->name = kname; | |
94b5d262 AV |
176 | len = strncpy_from_user(kname, filename, PATH_MAX); |
177 | if (unlikely(len < 0)) { | |
178 | __putname(kname); | |
179 | kfree(result); | |
180 | return ERR_PTR(len); | |
181 | } | |
182 | if (unlikely(len == PATH_MAX)) { | |
183 | __putname(kname); | |
184 | kfree(result); | |
185 | return ERR_PTR(-ENAMETOOLONG); | |
186 | } | |
7950e385 JL |
187 | } |
188 | ||
94b5d262 | 189 | result->refcnt = 1; |
3f9f0aa6 LT |
190 | /* The empty path is special. */ |
191 | if (unlikely(!len)) { | |
192 | if (empty) | |
4043cde8 | 193 | *empty = 1; |
94b5d262 AV |
194 | if (!(flags & LOOKUP_EMPTY)) { |
195 | putname(result); | |
196 | return ERR_PTR(-ENOENT); | |
197 | } | |
1da177e4 | 198 | } |
3f9f0aa6 | 199 | |
7950e385 | 200 | result->uptr = filename; |
c4ad8f98 | 201 | result->aname = NULL; |
7950e385 JL |
202 | audit_getname(result); |
203 | return result; | |
1da177e4 LT |
204 | } |
205 | ||
91a27b2a JL |
206 | struct filename * |
207 | getname(const char __user * filename) | |
f52e0c11 | 208 | { |
f7493e5d | 209 | return getname_flags(filename, 0, NULL); |
f52e0c11 AV |
210 | } |
211 | ||
c4ad8f98 LT |
212 | struct filename * |
213 | getname_kernel(const char * filename) | |
214 | { | |
215 | struct filename *result; | |
08518549 | 216 | int len = strlen(filename) + 1; |
c4ad8f98 LT |
217 | |
218 | result = __getname(); | |
219 | if (unlikely(!result)) | |
220 | return ERR_PTR(-ENOMEM); | |
221 | ||
08518549 | 222 | if (len <= EMBEDDED_NAME_MAX) { |
fd2f7cb5 | 223 | result->name = (char *)result->iname; |
08518549 | 224 | } else if (len <= PATH_MAX) { |
30ce4d19 | 225 | const size_t size = offsetof(struct filename, iname[1]); |
08518549 PM |
226 | struct filename *tmp; |
227 | ||
30ce4d19 | 228 | tmp = kmalloc(size, GFP_KERNEL); |
08518549 PM |
229 | if (unlikely(!tmp)) { |
230 | __putname(result); | |
231 | return ERR_PTR(-ENOMEM); | |
232 | } | |
233 | tmp->name = (char *)result; | |
08518549 PM |
234 | result = tmp; |
235 | } else { | |
236 | __putname(result); | |
237 | return ERR_PTR(-ENAMETOOLONG); | |
238 | } | |
239 | memcpy((char *)result->name, filename, len); | |
c4ad8f98 LT |
240 | result->uptr = NULL; |
241 | result->aname = NULL; | |
55422d0b | 242 | result->refcnt = 1; |
fd3522fd | 243 | audit_getname(result); |
c4ad8f98 | 244 | |
c4ad8f98 LT |
245 | return result; |
246 | } | |
247 | ||
91a27b2a | 248 | void putname(struct filename *name) |
1da177e4 | 249 | { |
55422d0b PM |
250 | BUG_ON(name->refcnt <= 0); |
251 | ||
252 | if (--name->refcnt > 0) | |
253 | return; | |
254 | ||
fd2f7cb5 | 255 | if (name->name != name->iname) { |
55422d0b PM |
256 | __putname(name->name); |
257 | kfree(name); | |
258 | } else | |
259 | __putname(name); | |
1da177e4 | 260 | } |
1da177e4 | 261 | |
e77819e5 LT |
262 | static int check_acl(struct inode *inode, int mask) |
263 | { | |
84635d68 | 264 | #ifdef CONFIG_FS_POSIX_ACL |
e77819e5 LT |
265 | struct posix_acl *acl; |
266 | ||
e77819e5 | 267 | if (mask & MAY_NOT_BLOCK) { |
3567866b AV |
268 | acl = get_cached_acl_rcu(inode, ACL_TYPE_ACCESS); |
269 | if (!acl) | |
e77819e5 | 270 | return -EAGAIN; |
3567866b | 271 | /* no ->get_acl() calls in RCU mode... */ |
b8a7a3a6 | 272 | if (is_uncached_acl(acl)) |
3567866b | 273 | return -ECHILD; |
63d72b93 | 274 | return posix_acl_permission(inode, acl, mask); |
e77819e5 LT |
275 | } |
276 | ||
2982baa2 CH |
277 | acl = get_acl(inode, ACL_TYPE_ACCESS); |
278 | if (IS_ERR(acl)) | |
279 | return PTR_ERR(acl); | |
e77819e5 LT |
280 | if (acl) { |
281 | int error = posix_acl_permission(inode, acl, mask); | |
282 | posix_acl_release(acl); | |
283 | return error; | |
284 | } | |
84635d68 | 285 | #endif |
e77819e5 LT |
286 | |
287 | return -EAGAIN; | |
288 | } | |
289 | ||
5909ccaa | 290 | /* |
5fc475b7 LT |
291 | * This does the basic UNIX permission checking. |
292 | * | |
293 | * Note that the POSIX ACL check cares about the MAY_NOT_BLOCK bit, | |
294 | * for RCU walking. | |
1da177e4 | 295 | */ |
7e40145e | 296 | static int acl_permission_check(struct inode *inode, int mask) |
1da177e4 | 297 | { |
26cf46be | 298 | unsigned int mode = inode->i_mode; |
1da177e4 | 299 | |
5fc475b7 LT |
300 | /* Are we the owner? If so, ACL's don't matter */ |
301 | if (likely(uid_eq(current_fsuid(), inode->i_uid))) { | |
302 | mask &= 7; | |
1da177e4 | 303 | mode >>= 6; |
5fc475b7 LT |
304 | return (mask & ~mode) ? -EACCES : 0; |
305 | } | |
1da177e4 | 306 | |
5fc475b7 LT |
307 | /* Do we have ACL's? */ |
308 | if (IS_POSIXACL(inode) && (mode & S_IRWXG)) { | |
309 | int error = check_acl(inode, mask); | |
310 | if (error != -EAGAIN) | |
311 | return error; | |
1da177e4 LT |
312 | } |
313 | ||
5fc475b7 LT |
314 | /* Only RWX matters for group/other mode bits */ |
315 | mask &= 7; | |
316 | ||
1da177e4 | 317 | /* |
5fc475b7 LT |
318 | * Are the group permissions different from |
319 | * the other permissions in the bits we care | |
320 | * about? Need to check group ownership if so. | |
1da177e4 | 321 | */ |
5fc475b7 LT |
322 | if (mask & (mode ^ (mode >> 3))) { |
323 | if (in_group_p(inode->i_gid)) | |
324 | mode >>= 3; | |
325 | } | |
326 | ||
327 | /* Bits in 'mode' clear that we require? */ | |
328 | return (mask & ~mode) ? -EACCES : 0; | |
5909ccaa LT |
329 | } |
330 | ||
331 | /** | |
b74c79e9 | 332 | * generic_permission - check for access rights on a Posix-like filesystem |
5909ccaa | 333 | * @inode: inode to check access rights for |
5fc475b7 LT |
334 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC, |
335 | * %MAY_NOT_BLOCK ...) | |
5909ccaa LT |
336 | * |
337 | * Used to check for read/write/execute permissions on a file. | |
338 | * We use "fsuid" for this, letting us set arbitrary permissions | |
339 | * for filesystem access without changing the "normal" uids which | |
b74c79e9 NP |
340 | * are used for other things. |
341 | * | |
342 | * generic_permission is rcu-walk aware. It returns -ECHILD in case an rcu-walk | |
343 | * request cannot be satisfied (eg. requires blocking or too much complexity). | |
344 | * It would then be called again in ref-walk mode. | |
5909ccaa | 345 | */ |
2830ba7f | 346 | int generic_permission(struct inode *inode, int mask) |
5909ccaa LT |
347 | { |
348 | int ret; | |
349 | ||
350 | /* | |
948409c7 | 351 | * Do the basic permission checks. |
5909ccaa | 352 | */ |
7e40145e | 353 | ret = acl_permission_check(inode, mask); |
5909ccaa LT |
354 | if (ret != -EACCES) |
355 | return ret; | |
1da177e4 | 356 | |
d594e7ec AV |
357 | if (S_ISDIR(inode->i_mode)) { |
358 | /* DACs are overridable for directories */ | |
d594e7ec | 359 | if (!(mask & MAY_WRITE)) |
0558c1bf | 360 | if (capable_wrt_inode_uidgid(&init_user_ns, inode, |
23adbe12 | 361 | CAP_DAC_READ_SEARCH)) |
d594e7ec | 362 | return 0; |
0558c1bf CB |
363 | if (capable_wrt_inode_uidgid(&init_user_ns, inode, |
364 | CAP_DAC_OVERRIDE)) | |
1da177e4 | 365 | return 0; |
2a4c2242 SS |
366 | return -EACCES; |
367 | } | |
1da177e4 LT |
368 | |
369 | /* | |
370 | * Searching includes executable on directories, else just read. | |
371 | */ | |
7ea66001 | 372 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
d594e7ec | 373 | if (mask == MAY_READ) |
0558c1bf CB |
374 | if (capable_wrt_inode_uidgid(&init_user_ns, inode, |
375 | CAP_DAC_READ_SEARCH)) | |
1da177e4 | 376 | return 0; |
2a4c2242 SS |
377 | /* |
378 | * Read/write DACs are always overridable. | |
379 | * Executable DACs are overridable when there is | |
380 | * at least one exec bit set. | |
381 | */ | |
382 | if (!(mask & MAY_EXEC) || (inode->i_mode & S_IXUGO)) | |
0558c1bf CB |
383 | if (capable_wrt_inode_uidgid(&init_user_ns, inode, |
384 | CAP_DAC_OVERRIDE)) | |
2a4c2242 | 385 | return 0; |
1da177e4 LT |
386 | |
387 | return -EACCES; | |
388 | } | |
4d359507 | 389 | EXPORT_SYMBOL(generic_permission); |
1da177e4 | 390 | |
3ddcd056 LT |
391 | /* |
392 | * We _really_ want to just do "generic_permission()" without | |
393 | * even looking at the inode->i_op values. So we keep a cache | |
394 | * flag in inode->i_opflags, that says "this has not special | |
395 | * permission function, use the fast case". | |
396 | */ | |
397 | static inline int do_inode_permission(struct inode *inode, int mask) | |
398 | { | |
399 | if (unlikely(!(inode->i_opflags & IOP_FASTPERM))) { | |
400 | if (likely(inode->i_op->permission)) | |
401 | return inode->i_op->permission(inode, mask); | |
402 | ||
403 | /* This gets set once for the inode lifetime */ | |
404 | spin_lock(&inode->i_lock); | |
405 | inode->i_opflags |= IOP_FASTPERM; | |
406 | spin_unlock(&inode->i_lock); | |
407 | } | |
408 | return generic_permission(inode, mask); | |
409 | } | |
410 | ||
0bdaea90 DH |
411 | /** |
412 | * sb_permission - Check superblock-level permissions | |
413 | * @sb: Superblock of inode to check permission on | |
55852635 | 414 | * @inode: Inode to check permission on |
0bdaea90 DH |
415 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) |
416 | * | |
417 | * Separate out file-system wide checks from inode-specific permission checks. | |
418 | */ | |
419 | static int sb_permission(struct super_block *sb, struct inode *inode, int mask) | |
420 | { | |
421 | if (unlikely(mask & MAY_WRITE)) { | |
422 | umode_t mode = inode->i_mode; | |
423 | ||
424 | /* Nobody gets write access to a read-only fs. */ | |
bc98a42c | 425 | if (sb_rdonly(sb) && (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) |
0bdaea90 DH |
426 | return -EROFS; |
427 | } | |
428 | return 0; | |
429 | } | |
430 | ||
431 | /** | |
432 | * inode_permission - Check for access rights to a given inode | |
433 | * @inode: Inode to check permission on | |
434 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
435 | * | |
436 | * Check for read/write/execute permissions on an inode. We use fs[ug]id for | |
437 | * this, letting us set arbitrary permissions for filesystem access without | |
438 | * changing the "normal" UIDs which are used for other things. | |
439 | * | |
440 | * When checking for MAY_APPEND, MAY_WRITE must also be set in @mask. | |
441 | */ | |
442 | int inode_permission(struct inode *inode, int mask) | |
443 | { | |
444 | int retval; | |
445 | ||
446 | retval = sb_permission(inode->i_sb, inode, mask); | |
447 | if (retval) | |
448 | return retval; | |
4bfd054a EB |
449 | |
450 | if (unlikely(mask & MAY_WRITE)) { | |
451 | /* | |
452 | * Nobody gets write access to an immutable file. | |
453 | */ | |
454 | if (IS_IMMUTABLE(inode)) | |
455 | return -EPERM; | |
456 | ||
457 | /* | |
458 | * Updating mtime will likely cause i_uid and i_gid to be | |
459 | * written back improperly if their true value is unknown | |
460 | * to the vfs. | |
461 | */ | |
462 | if (HAS_UNMAPPED_ID(inode)) | |
463 | return -EACCES; | |
464 | } | |
465 | ||
466 | retval = do_inode_permission(inode, mask); | |
467 | if (retval) | |
468 | return retval; | |
469 | ||
470 | retval = devcgroup_inode_permission(inode, mask); | |
471 | if (retval) | |
472 | return retval; | |
473 | ||
474 | return security_inode_permission(inode, mask); | |
0bdaea90 | 475 | } |
4d359507 | 476 | EXPORT_SYMBOL(inode_permission); |
0bdaea90 | 477 | |
5dd784d0 JB |
478 | /** |
479 | * path_get - get a reference to a path | |
480 | * @path: path to get the reference to | |
481 | * | |
482 | * Given a path increment the reference count to the dentry and the vfsmount. | |
483 | */ | |
dcf787f3 | 484 | void path_get(const struct path *path) |
5dd784d0 JB |
485 | { |
486 | mntget(path->mnt); | |
487 | dget(path->dentry); | |
488 | } | |
489 | EXPORT_SYMBOL(path_get); | |
490 | ||
1d957f9b JB |
491 | /** |
492 | * path_put - put a reference to a path | |
493 | * @path: path to put the reference to | |
494 | * | |
495 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
496 | */ | |
dcf787f3 | 497 | void path_put(const struct path *path) |
1da177e4 | 498 | { |
1d957f9b JB |
499 | dput(path->dentry); |
500 | mntput(path->mnt); | |
1da177e4 | 501 | } |
1d957f9b | 502 | EXPORT_SYMBOL(path_put); |
1da177e4 | 503 | |
894bc8c4 | 504 | #define EMBEDDED_LEVELS 2 |
1f55a6ec AV |
505 | struct nameidata { |
506 | struct path path; | |
1cf2665b | 507 | struct qstr last; |
1f55a6ec AV |
508 | struct path root; |
509 | struct inode *inode; /* path.dentry.d_inode */ | |
510 | unsigned int flags; | |
ab87f9a5 | 511 | unsigned seq, m_seq, r_seq; |
1f55a6ec AV |
512 | int last_type; |
513 | unsigned depth; | |
756daf26 | 514 | int total_link_count; |
697fc6ca AV |
515 | struct saved { |
516 | struct path link; | |
fceef393 | 517 | struct delayed_call done; |
697fc6ca | 518 | const char *name; |
0450b2d1 | 519 | unsigned seq; |
894bc8c4 | 520 | } *stack, internal[EMBEDDED_LEVELS]; |
9883d185 AV |
521 | struct filename *name; |
522 | struct nameidata *saved; | |
523 | unsigned root_seq; | |
524 | int dfd; | |
0f705953 AV |
525 | kuid_t dir_uid; |
526 | umode_t dir_mode; | |
3859a271 | 527 | } __randomize_layout; |
1f55a6ec | 528 | |
9883d185 | 529 | static void set_nameidata(struct nameidata *p, int dfd, struct filename *name) |
894bc8c4 | 530 | { |
756daf26 N |
531 | struct nameidata *old = current->nameidata; |
532 | p->stack = p->internal; | |
c8a53ee5 AV |
533 | p->dfd = dfd; |
534 | p->name = name; | |
756daf26 | 535 | p->total_link_count = old ? old->total_link_count : 0; |
9883d185 | 536 | p->saved = old; |
756daf26 | 537 | current->nameidata = p; |
894bc8c4 AV |
538 | } |
539 | ||
9883d185 | 540 | static void restore_nameidata(void) |
894bc8c4 | 541 | { |
9883d185 | 542 | struct nameidata *now = current->nameidata, *old = now->saved; |
756daf26 N |
543 | |
544 | current->nameidata = old; | |
545 | if (old) | |
546 | old->total_link_count = now->total_link_count; | |
e1a63bbc | 547 | if (now->stack != now->internal) |
756daf26 | 548 | kfree(now->stack); |
894bc8c4 AV |
549 | } |
550 | ||
60ef60c7 | 551 | static bool nd_alloc_stack(struct nameidata *nd) |
894bc8c4 | 552 | { |
bc40aee0 AV |
553 | struct saved *p; |
554 | ||
60ef60c7 AV |
555 | p= kmalloc_array(MAXSYMLINKS, sizeof(struct saved), |
556 | nd->flags & LOOKUP_RCU ? GFP_ATOMIC : GFP_KERNEL); | |
557 | if (unlikely(!p)) | |
558 | return false; | |
894bc8c4 AV |
559 | memcpy(p, nd->internal, sizeof(nd->internal)); |
560 | nd->stack = p; | |
60ef60c7 | 561 | return true; |
894bc8c4 AV |
562 | } |
563 | ||
397d425d | 564 | /** |
6b03f7ed | 565 | * path_connected - Verify that a dentry is below mnt.mnt_root |
397d425d EB |
566 | * |
567 | * Rename can sometimes move a file or directory outside of a bind | |
568 | * mount, path_connected allows those cases to be detected. | |
569 | */ | |
6b03f7ed | 570 | static bool path_connected(struct vfsmount *mnt, struct dentry *dentry) |
397d425d | 571 | { |
95dd7758 | 572 | struct super_block *sb = mnt->mnt_sb; |
397d425d | 573 | |
402dd2cf CH |
574 | /* Bind mounts can have disconnected paths */ |
575 | if (mnt->mnt_root == sb->s_root) | |
397d425d EB |
576 | return true; |
577 | ||
6b03f7ed | 578 | return is_subdir(dentry, mnt->mnt_root); |
397d425d EB |
579 | } |
580 | ||
7973387a AV |
581 | static void drop_links(struct nameidata *nd) |
582 | { | |
583 | int i = nd->depth; | |
584 | while (i--) { | |
585 | struct saved *last = nd->stack + i; | |
fceef393 AV |
586 | do_delayed_call(&last->done); |
587 | clear_delayed_call(&last->done); | |
7973387a AV |
588 | } |
589 | } | |
590 | ||
591 | static void terminate_walk(struct nameidata *nd) | |
592 | { | |
593 | drop_links(nd); | |
594 | if (!(nd->flags & LOOKUP_RCU)) { | |
595 | int i; | |
596 | path_put(&nd->path); | |
597 | for (i = 0; i < nd->depth; i++) | |
598 | path_put(&nd->stack[i].link); | |
84a2bd39 | 599 | if (nd->flags & LOOKUP_ROOT_GRABBED) { |
102b8af2 | 600 | path_put(&nd->root); |
84a2bd39 | 601 | nd->flags &= ~LOOKUP_ROOT_GRABBED; |
102b8af2 | 602 | } |
7973387a AV |
603 | } else { |
604 | nd->flags &= ~LOOKUP_RCU; | |
7973387a AV |
605 | rcu_read_unlock(); |
606 | } | |
607 | nd->depth = 0; | |
608 | } | |
609 | ||
610 | /* path_put is needed afterwards regardless of success or failure */ | |
2aa38470 | 611 | static bool __legitimize_path(struct path *path, unsigned seq, unsigned mseq) |
7973387a | 612 | { |
2aa38470 | 613 | int res = __legitimize_mnt(path->mnt, mseq); |
7973387a AV |
614 | if (unlikely(res)) { |
615 | if (res > 0) | |
616 | path->mnt = NULL; | |
617 | path->dentry = NULL; | |
618 | return false; | |
619 | } | |
620 | if (unlikely(!lockref_get_not_dead(&path->dentry->d_lockref))) { | |
621 | path->dentry = NULL; | |
622 | return false; | |
623 | } | |
624 | return !read_seqcount_retry(&path->dentry->d_seq, seq); | |
625 | } | |
626 | ||
2aa38470 AV |
627 | static inline bool legitimize_path(struct nameidata *nd, |
628 | struct path *path, unsigned seq) | |
629 | { | |
5bd73286 | 630 | return __legitimize_path(path, seq, nd->m_seq); |
2aa38470 AV |
631 | } |
632 | ||
7973387a AV |
633 | static bool legitimize_links(struct nameidata *nd) |
634 | { | |
635 | int i; | |
636 | for (i = 0; i < nd->depth; i++) { | |
637 | struct saved *last = nd->stack + i; | |
638 | if (unlikely(!legitimize_path(nd, &last->link, last->seq))) { | |
639 | drop_links(nd); | |
640 | nd->depth = i + 1; | |
641 | return false; | |
642 | } | |
643 | } | |
644 | return true; | |
645 | } | |
646 | ||
ee594bff AV |
647 | static bool legitimize_root(struct nameidata *nd) |
648 | { | |
adb21d2b AS |
649 | /* |
650 | * For scoped-lookups (where nd->root has been zeroed), we need to | |
651 | * restart the whole lookup from scratch -- because set_root() is wrong | |
652 | * for these lookups (nd->dfd is the root, not the filesystem root). | |
653 | */ | |
654 | if (!nd->root.mnt && (nd->flags & LOOKUP_IS_SCOPED)) | |
655 | return false; | |
656 | /* Nothing to do if nd->root is zero or is managed by the VFS user. */ | |
ee594bff AV |
657 | if (!nd->root.mnt || (nd->flags & LOOKUP_ROOT)) |
658 | return true; | |
84a2bd39 | 659 | nd->flags |= LOOKUP_ROOT_GRABBED; |
ee594bff AV |
660 | return legitimize_path(nd, &nd->root, nd->root_seq); |
661 | } | |
662 | ||
19660af7 | 663 | /* |
31e6b01f | 664 | * Path walking has 2 modes, rcu-walk and ref-walk (see |
19660af7 AV |
665 | * Documentation/filesystems/path-lookup.txt). In situations when we can't |
666 | * continue in RCU mode, we attempt to drop out of rcu-walk mode and grab | |
57e3715c | 667 | * normal reference counts on dentries and vfsmounts to transition to ref-walk |
19660af7 AV |
668 | * mode. Refcounts are grabbed at the last known good point before rcu-walk |
669 | * got stuck, so ref-walk may continue from there. If this is not successful | |
670 | * (eg. a seqcount has changed), then failure is returned and it's up to caller | |
671 | * to restart the path walk from the beginning in ref-walk mode. | |
31e6b01f | 672 | */ |
31e6b01f NP |
673 | |
674 | /** | |
19660af7 AV |
675 | * unlazy_walk - try to switch to ref-walk mode. |
676 | * @nd: nameidata pathwalk data | |
39191628 | 677 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f | 678 | * |
4675ac39 AV |
679 | * unlazy_walk attempts to legitimize the current nd->path and nd->root |
680 | * for ref-walk mode. | |
681 | * Must be called from rcu-walk context. | |
7973387a AV |
682 | * Nothing should touch nameidata between unlazy_walk() failure and |
683 | * terminate_walk(). | |
31e6b01f | 684 | */ |
4675ac39 | 685 | static int unlazy_walk(struct nameidata *nd) |
31e6b01f | 686 | { |
31e6b01f NP |
687 | struct dentry *parent = nd->path.dentry; |
688 | ||
689 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
e5c832d5 | 690 | |
4675ac39 AV |
691 | nd->flags &= ~LOOKUP_RCU; |
692 | if (unlikely(!legitimize_links(nd))) | |
4675ac39 | 693 | goto out1; |
84a2bd39 AV |
694 | if (unlikely(!legitimize_path(nd, &nd->path, nd->seq))) |
695 | goto out; | |
ee594bff AV |
696 | if (unlikely(!legitimize_root(nd))) |
697 | goto out; | |
4675ac39 AV |
698 | rcu_read_unlock(); |
699 | BUG_ON(nd->inode != parent->d_inode); | |
700 | return 0; | |
701 | ||
84a2bd39 | 702 | out1: |
4675ac39 AV |
703 | nd->path.mnt = NULL; |
704 | nd->path.dentry = NULL; | |
4675ac39 AV |
705 | out: |
706 | rcu_read_unlock(); | |
707 | return -ECHILD; | |
708 | } | |
709 | ||
710 | /** | |
711 | * unlazy_child - try to switch to ref-walk mode. | |
712 | * @nd: nameidata pathwalk data | |
713 | * @dentry: child of nd->path.dentry | |
714 | * @seq: seq number to check dentry against | |
715 | * Returns: 0 on success, -ECHILD on failure | |
716 | * | |
717 | * unlazy_child attempts to legitimize the current nd->path, nd->root and dentry | |
718 | * for ref-walk mode. @dentry must be a path found by a do_lookup call on | |
719 | * @nd. Must be called from rcu-walk context. | |
720 | * Nothing should touch nameidata between unlazy_child() failure and | |
721 | * terminate_walk(). | |
722 | */ | |
723 | static int unlazy_child(struct nameidata *nd, struct dentry *dentry, unsigned seq) | |
724 | { | |
725 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
726 | ||
e5c832d5 | 727 | nd->flags &= ~LOOKUP_RCU; |
7973387a AV |
728 | if (unlikely(!legitimize_links(nd))) |
729 | goto out2; | |
730 | if (unlikely(!legitimize_mnt(nd->path.mnt, nd->m_seq))) | |
731 | goto out2; | |
4675ac39 | 732 | if (unlikely(!lockref_get_not_dead(&nd->path.dentry->d_lockref))) |
7973387a | 733 | goto out1; |
48a066e7 | 734 | |
15570086 | 735 | /* |
4675ac39 AV |
736 | * We need to move both the parent and the dentry from the RCU domain |
737 | * to be properly refcounted. And the sequence number in the dentry | |
738 | * validates *both* dentry counters, since we checked the sequence | |
739 | * number of the parent after we got the child sequence number. So we | |
740 | * know the parent must still be valid if the child sequence number is | |
15570086 | 741 | */ |
4675ac39 AV |
742 | if (unlikely(!lockref_get_not_dead(&dentry->d_lockref))) |
743 | goto out; | |
84a2bd39 AV |
744 | if (unlikely(read_seqcount_retry(&dentry->d_seq, seq))) |
745 | goto out_dput; | |
e5c832d5 LT |
746 | /* |
747 | * Sequence counts matched. Now make sure that the root is | |
748 | * still valid and get it if required. | |
749 | */ | |
84a2bd39 AV |
750 | if (unlikely(!legitimize_root(nd))) |
751 | goto out_dput; | |
8b61e74f | 752 | rcu_read_unlock(); |
31e6b01f | 753 | return 0; |
19660af7 | 754 | |
7973387a AV |
755 | out2: |
756 | nd->path.mnt = NULL; | |
757 | out1: | |
758 | nd->path.dentry = NULL; | |
e5c832d5 | 759 | out: |
8b61e74f | 760 | rcu_read_unlock(); |
84a2bd39 AV |
761 | return -ECHILD; |
762 | out_dput: | |
763 | rcu_read_unlock(); | |
764 | dput(dentry); | |
31e6b01f NP |
765 | return -ECHILD; |
766 | } | |
767 | ||
4ce16ef3 | 768 | static inline int d_revalidate(struct dentry *dentry, unsigned int flags) |
34286d66 | 769 | { |
a89f8337 AV |
770 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) |
771 | return dentry->d_op->d_revalidate(dentry, flags); | |
772 | else | |
773 | return 1; | |
34286d66 NP |
774 | } |
775 | ||
9f1fafee AV |
776 | /** |
777 | * complete_walk - successful completion of path walk | |
778 | * @nd: pointer nameidata | |
39159de2 | 779 | * |
9f1fafee AV |
780 | * If we had been in RCU mode, drop out of it and legitimize nd->path. |
781 | * Revalidate the final result, unless we'd already done that during | |
782 | * the path walk or the filesystem doesn't ask for it. Return 0 on | |
783 | * success, -error on failure. In case of failure caller does not | |
784 | * need to drop nd->path. | |
39159de2 | 785 | */ |
9f1fafee | 786 | static int complete_walk(struct nameidata *nd) |
39159de2 | 787 | { |
16c2cd71 | 788 | struct dentry *dentry = nd->path.dentry; |
39159de2 | 789 | int status; |
39159de2 | 790 | |
9f1fafee | 791 | if (nd->flags & LOOKUP_RCU) { |
adb21d2b AS |
792 | /* |
793 | * We don't want to zero nd->root for scoped-lookups or | |
794 | * externally-managed nd->root. | |
795 | */ | |
796 | if (!(nd->flags & (LOOKUP_ROOT | LOOKUP_IS_SCOPED))) | |
9f1fafee | 797 | nd->root.mnt = NULL; |
4675ac39 | 798 | if (unlikely(unlazy_walk(nd))) |
9f1fafee | 799 | return -ECHILD; |
9f1fafee AV |
800 | } |
801 | ||
adb21d2b AS |
802 | if (unlikely(nd->flags & LOOKUP_IS_SCOPED)) { |
803 | /* | |
804 | * While the guarantee of LOOKUP_IS_SCOPED is (roughly) "don't | |
805 | * ever step outside the root during lookup" and should already | |
806 | * be guaranteed by the rest of namei, we want to avoid a namei | |
807 | * BUG resulting in userspace being given a path that was not | |
808 | * scoped within the root at some point during the lookup. | |
809 | * | |
810 | * So, do a final sanity-check to make sure that in the | |
811 | * worst-case scenario (a complete bypass of LOOKUP_IS_SCOPED) | |
812 | * we won't silently return an fd completely outside of the | |
813 | * requested root to userspace. | |
814 | * | |
815 | * Userspace could move the path outside the root after this | |
816 | * check, but as discussed elsewhere this is not a concern (the | |
817 | * resolved file was inside the root at some point). | |
818 | */ | |
819 | if (!path_is_under(&nd->path, &nd->root)) | |
820 | return -EXDEV; | |
821 | } | |
822 | ||
16c2cd71 AV |
823 | if (likely(!(nd->flags & LOOKUP_JUMPED))) |
824 | return 0; | |
825 | ||
ecf3d1f1 | 826 | if (likely(!(dentry->d_flags & DCACHE_OP_WEAK_REVALIDATE))) |
39159de2 JL |
827 | return 0; |
828 | ||
ecf3d1f1 | 829 | status = dentry->d_op->d_weak_revalidate(dentry, nd->flags); |
39159de2 JL |
830 | if (status > 0) |
831 | return 0; | |
832 | ||
16c2cd71 | 833 | if (!status) |
39159de2 | 834 | status = -ESTALE; |
16c2cd71 | 835 | |
39159de2 JL |
836 | return status; |
837 | } | |
838 | ||
740a1678 | 839 | static int set_root(struct nameidata *nd) |
31e6b01f | 840 | { |
7bd88377 | 841 | struct fs_struct *fs = current->fs; |
c28cc364 | 842 | |
adb21d2b AS |
843 | /* |
844 | * Jumping to the real root in a scoped-lookup is a BUG in namei, but we | |
845 | * still have to ensure it doesn't happen because it will cause a breakout | |
846 | * from the dirfd. | |
847 | */ | |
848 | if (WARN_ON(nd->flags & LOOKUP_IS_SCOPED)) | |
849 | return -ENOTRECOVERABLE; | |
850 | ||
9e6697e2 AV |
851 | if (nd->flags & LOOKUP_RCU) { |
852 | unsigned seq; | |
853 | ||
854 | do { | |
855 | seq = read_seqcount_begin(&fs->seq); | |
856 | nd->root = fs->root; | |
857 | nd->root_seq = __read_seqcount_begin(&nd->root.dentry->d_seq); | |
858 | } while (read_seqcount_retry(&fs->seq, seq)); | |
859 | } else { | |
860 | get_fs_root(fs, &nd->root); | |
84a2bd39 | 861 | nd->flags |= LOOKUP_ROOT_GRABBED; |
9e6697e2 | 862 | } |
740a1678 | 863 | return 0; |
31e6b01f NP |
864 | } |
865 | ||
248fb5b9 AV |
866 | static int nd_jump_root(struct nameidata *nd) |
867 | { | |
adb21d2b AS |
868 | if (unlikely(nd->flags & LOOKUP_BENEATH)) |
869 | return -EXDEV; | |
72ba2929 AS |
870 | if (unlikely(nd->flags & LOOKUP_NO_XDEV)) { |
871 | /* Absolute path arguments to path_init() are allowed. */ | |
872 | if (nd->path.mnt != NULL && nd->path.mnt != nd->root.mnt) | |
873 | return -EXDEV; | |
874 | } | |
740a1678 AS |
875 | if (!nd->root.mnt) { |
876 | int error = set_root(nd); | |
877 | if (error) | |
878 | return error; | |
879 | } | |
248fb5b9 AV |
880 | if (nd->flags & LOOKUP_RCU) { |
881 | struct dentry *d; | |
882 | nd->path = nd->root; | |
883 | d = nd->path.dentry; | |
884 | nd->inode = d->d_inode; | |
885 | nd->seq = nd->root_seq; | |
886 | if (unlikely(read_seqcount_retry(&d->d_seq, nd->seq))) | |
887 | return -ECHILD; | |
888 | } else { | |
889 | path_put(&nd->path); | |
890 | nd->path = nd->root; | |
891 | path_get(&nd->path); | |
892 | nd->inode = nd->path.dentry->d_inode; | |
893 | } | |
894 | nd->flags |= LOOKUP_JUMPED; | |
895 | return 0; | |
896 | } | |
897 | ||
b5fb63c1 | 898 | /* |
6b255391 | 899 | * Helper to directly jump to a known parsed path from ->get_link, |
b5fb63c1 CH |
900 | * caller must have taken a reference to path beforehand. |
901 | */ | |
1bc82070 | 902 | int nd_jump_link(struct path *path) |
b5fb63c1 | 903 | { |
4b99d499 | 904 | int error = -ELOOP; |
6e77137b | 905 | struct nameidata *nd = current->nameidata; |
b5fb63c1 | 906 | |
4b99d499 AS |
907 | if (unlikely(nd->flags & LOOKUP_NO_MAGICLINKS)) |
908 | goto err; | |
909 | ||
72ba2929 AS |
910 | error = -EXDEV; |
911 | if (unlikely(nd->flags & LOOKUP_NO_XDEV)) { | |
912 | if (nd->path.mnt != path->mnt) | |
913 | goto err; | |
914 | } | |
adb21d2b AS |
915 | /* Not currently safe for scoped-lookups. */ |
916 | if (unlikely(nd->flags & LOOKUP_IS_SCOPED)) | |
917 | goto err; | |
72ba2929 | 918 | |
4b99d499 | 919 | path_put(&nd->path); |
b5fb63c1 CH |
920 | nd->path = *path; |
921 | nd->inode = nd->path.dentry->d_inode; | |
922 | nd->flags |= LOOKUP_JUMPED; | |
1bc82070 | 923 | return 0; |
4b99d499 AS |
924 | |
925 | err: | |
926 | path_put(path); | |
927 | return error; | |
b5fb63c1 CH |
928 | } |
929 | ||
b9ff4429 | 930 | static inline void put_link(struct nameidata *nd) |
574197e0 | 931 | { |
21c3003d | 932 | struct saved *last = nd->stack + --nd->depth; |
fceef393 | 933 | do_delayed_call(&last->done); |
6548fae2 AV |
934 | if (!(nd->flags & LOOKUP_RCU)) |
935 | path_put(&last->link); | |
574197e0 AV |
936 | } |
937 | ||
561ec64a LT |
938 | int sysctl_protected_symlinks __read_mostly = 0; |
939 | int sysctl_protected_hardlinks __read_mostly = 0; | |
30aba665 SM |
940 | int sysctl_protected_fifos __read_mostly; |
941 | int sysctl_protected_regular __read_mostly; | |
800179c9 KC |
942 | |
943 | /** | |
944 | * may_follow_link - Check symlink following for unsafe situations | |
55852635 | 945 | * @nd: nameidata pathwalk data |
800179c9 KC |
946 | * |
947 | * In the case of the sysctl_protected_symlinks sysctl being enabled, | |
948 | * CAP_DAC_OVERRIDE needs to be specifically ignored if the symlink is | |
949 | * in a sticky world-writable directory. This is to protect privileged | |
950 | * processes from failing races against path names that may change out | |
951 | * from under them by way of other users creating malicious symlinks. | |
952 | * It will permit symlinks to be followed only when outside a sticky | |
953 | * world-writable directory, or when the uid of the symlink and follower | |
954 | * match, or when the directory owner matches the symlink's owner. | |
955 | * | |
956 | * Returns 0 if following the symlink is allowed, -ve on error. | |
957 | */ | |
ad6cc4c3 | 958 | static inline int may_follow_link(struct nameidata *nd, const struct inode *inode) |
800179c9 | 959 | { |
800179c9 KC |
960 | if (!sysctl_protected_symlinks) |
961 | return 0; | |
962 | ||
963 | /* Allowed if owner and follower match. */ | |
81abe27b | 964 | if (uid_eq(current_cred()->fsuid, inode->i_uid)) |
800179c9 KC |
965 | return 0; |
966 | ||
967 | /* Allowed if parent directory not sticky and world-writable. */ | |
0f705953 | 968 | if ((nd->dir_mode & (S_ISVTX|S_IWOTH)) != (S_ISVTX|S_IWOTH)) |
800179c9 KC |
969 | return 0; |
970 | ||
971 | /* Allowed if parent directory and link owner match. */ | |
0f705953 | 972 | if (uid_valid(nd->dir_uid) && uid_eq(nd->dir_uid, inode->i_uid)) |
800179c9 KC |
973 | return 0; |
974 | ||
31956502 AV |
975 | if (nd->flags & LOOKUP_RCU) |
976 | return -ECHILD; | |
977 | ||
ea841baf | 978 | audit_inode(nd->name, nd->stack[0].link.dentry, 0); |
245d7369 | 979 | audit_log_path_denied(AUDIT_ANOM_LINK, "follow_link"); |
800179c9 KC |
980 | return -EACCES; |
981 | } | |
982 | ||
983 | /** | |
984 | * safe_hardlink_source - Check for safe hardlink conditions | |
985 | * @inode: the source inode to hardlink from | |
986 | * | |
987 | * Return false if at least one of the following conditions: | |
988 | * - inode is not a regular file | |
989 | * - inode is setuid | |
990 | * - inode is setgid and group-exec | |
991 | * - access failure for read and write | |
992 | * | |
993 | * Otherwise returns true. | |
994 | */ | |
995 | static bool safe_hardlink_source(struct inode *inode) | |
996 | { | |
997 | umode_t mode = inode->i_mode; | |
998 | ||
999 | /* Special files should not get pinned to the filesystem. */ | |
1000 | if (!S_ISREG(mode)) | |
1001 | return false; | |
1002 | ||
1003 | /* Setuid files should not get pinned to the filesystem. */ | |
1004 | if (mode & S_ISUID) | |
1005 | return false; | |
1006 | ||
1007 | /* Executable setgid files should not get pinned to the filesystem. */ | |
1008 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) | |
1009 | return false; | |
1010 | ||
1011 | /* Hardlinking to unreadable or unwritable sources is dangerous. */ | |
1012 | if (inode_permission(inode, MAY_READ | MAY_WRITE)) | |
1013 | return false; | |
1014 | ||
1015 | return true; | |
1016 | } | |
1017 | ||
1018 | /** | |
1019 | * may_linkat - Check permissions for creating a hardlink | |
1020 | * @link: the source to hardlink from | |
1021 | * | |
1022 | * Block hardlink when all of: | |
1023 | * - sysctl_protected_hardlinks enabled | |
1024 | * - fsuid does not match inode | |
1025 | * - hardlink source is unsafe (see safe_hardlink_source() above) | |
f2ca3796 | 1026 | * - not CAP_FOWNER in a namespace with the inode owner uid mapped |
800179c9 KC |
1027 | * |
1028 | * Returns 0 if successful, -ve on error. | |
1029 | */ | |
812931d6 | 1030 | int may_linkat(struct path *link) |
800179c9 | 1031 | { |
593d1ce8 EB |
1032 | struct inode *inode = link->dentry->d_inode; |
1033 | ||
1034 | /* Inode writeback is not safe when the uid or gid are invalid. */ | |
1035 | if (!uid_valid(inode->i_uid) || !gid_valid(inode->i_gid)) | |
1036 | return -EOVERFLOW; | |
800179c9 KC |
1037 | |
1038 | if (!sysctl_protected_hardlinks) | |
1039 | return 0; | |
1040 | ||
800179c9 KC |
1041 | /* Source inode owner (or CAP_FOWNER) can hardlink all they like, |
1042 | * otherwise, it must be a safe source. | |
1043 | */ | |
cc658db4 | 1044 | if (safe_hardlink_source(inode) || inode_owner_or_capable(inode)) |
800179c9 KC |
1045 | return 0; |
1046 | ||
245d7369 | 1047 | audit_log_path_denied(AUDIT_ANOM_LINK, "linkat"); |
800179c9 KC |
1048 | return -EPERM; |
1049 | } | |
1050 | ||
30aba665 SM |
1051 | /** |
1052 | * may_create_in_sticky - Check whether an O_CREAT open in a sticky directory | |
1053 | * should be allowed, or not, on files that already | |
1054 | * exist. | |
d0cb5018 AV |
1055 | * @dir_mode: mode bits of directory |
1056 | * @dir_uid: owner of directory | |
30aba665 SM |
1057 | * @inode: the inode of the file to open |
1058 | * | |
1059 | * Block an O_CREAT open of a FIFO (or a regular file) when: | |
1060 | * - sysctl_protected_fifos (or sysctl_protected_regular) is enabled | |
1061 | * - the file already exists | |
1062 | * - we are in a sticky directory | |
1063 | * - we don't own the file | |
1064 | * - the owner of the directory doesn't own the file | |
1065 | * - the directory is world writable | |
1066 | * If the sysctl_protected_fifos (or sysctl_protected_regular) is set to 2 | |
1067 | * the directory doesn't have to be world writable: being group writable will | |
1068 | * be enough. | |
1069 | * | |
1070 | * Returns 0 if the open is allowed, -ve on error. | |
1071 | */ | |
d0cb5018 | 1072 | static int may_create_in_sticky(umode_t dir_mode, kuid_t dir_uid, |
30aba665 SM |
1073 | struct inode * const inode) |
1074 | { | |
1075 | if ((!sysctl_protected_fifos && S_ISFIFO(inode->i_mode)) || | |
1076 | (!sysctl_protected_regular && S_ISREG(inode->i_mode)) || | |
d0cb5018 AV |
1077 | likely(!(dir_mode & S_ISVTX)) || |
1078 | uid_eq(inode->i_uid, dir_uid) || | |
30aba665 SM |
1079 | uid_eq(current_fsuid(), inode->i_uid)) |
1080 | return 0; | |
1081 | ||
d0cb5018 AV |
1082 | if (likely(dir_mode & 0002) || |
1083 | (dir_mode & 0020 && | |
30aba665 SM |
1084 | ((sysctl_protected_fifos >= 2 && S_ISFIFO(inode->i_mode)) || |
1085 | (sysctl_protected_regular >= 2 && S_ISREG(inode->i_mode))))) { | |
245d7369 KC |
1086 | const char *operation = S_ISFIFO(inode->i_mode) ? |
1087 | "sticky_create_fifo" : | |
1088 | "sticky_create_regular"; | |
1089 | audit_log_path_denied(AUDIT_ANOM_CREAT, operation); | |
30aba665 SM |
1090 | return -EACCES; |
1091 | } | |
1092 | return 0; | |
1093 | } | |
1094 | ||
f015f126 DH |
1095 | /* |
1096 | * follow_up - Find the mountpoint of path's vfsmount | |
1097 | * | |
1098 | * Given a path, find the mountpoint of its source file system. | |
1099 | * Replace @path with the path of the mountpoint in the parent mount. | |
1100 | * Up is towards /. | |
1101 | * | |
1102 | * Return 1 if we went up a level and 0 if we were already at the | |
1103 | * root. | |
1104 | */ | |
bab77ebf | 1105 | int follow_up(struct path *path) |
1da177e4 | 1106 | { |
0714a533 AV |
1107 | struct mount *mnt = real_mount(path->mnt); |
1108 | struct mount *parent; | |
1da177e4 | 1109 | struct dentry *mountpoint; |
99b7db7b | 1110 | |
48a066e7 | 1111 | read_seqlock_excl(&mount_lock); |
0714a533 | 1112 | parent = mnt->mnt_parent; |
3c0a6163 | 1113 | if (parent == mnt) { |
48a066e7 | 1114 | read_sequnlock_excl(&mount_lock); |
1da177e4 LT |
1115 | return 0; |
1116 | } | |
0714a533 | 1117 | mntget(&parent->mnt); |
a73324da | 1118 | mountpoint = dget(mnt->mnt_mountpoint); |
48a066e7 | 1119 | read_sequnlock_excl(&mount_lock); |
bab77ebf AV |
1120 | dput(path->dentry); |
1121 | path->dentry = mountpoint; | |
1122 | mntput(path->mnt); | |
0714a533 | 1123 | path->mnt = &parent->mnt; |
1da177e4 LT |
1124 | return 1; |
1125 | } | |
4d359507 | 1126 | EXPORT_SYMBOL(follow_up); |
1da177e4 | 1127 | |
7ef482fa AV |
1128 | static bool choose_mountpoint_rcu(struct mount *m, const struct path *root, |
1129 | struct path *path, unsigned *seqp) | |
1130 | { | |
1131 | while (mnt_has_parent(m)) { | |
1132 | struct dentry *mountpoint = m->mnt_mountpoint; | |
1133 | ||
1134 | m = m->mnt_parent; | |
1135 | if (unlikely(root->dentry == mountpoint && | |
1136 | root->mnt == &m->mnt)) | |
1137 | break; | |
1138 | if (mountpoint != m->mnt.mnt_root) { | |
1139 | path->mnt = &m->mnt; | |
1140 | path->dentry = mountpoint; | |
1141 | *seqp = read_seqcount_begin(&mountpoint->d_seq); | |
1142 | return true; | |
1143 | } | |
1144 | } | |
1145 | return false; | |
1146 | } | |
1147 | ||
2aa38470 AV |
1148 | static bool choose_mountpoint(struct mount *m, const struct path *root, |
1149 | struct path *path) | |
1150 | { | |
1151 | bool found; | |
1152 | ||
1153 | rcu_read_lock(); | |
1154 | while (1) { | |
1155 | unsigned seq, mseq = read_seqbegin(&mount_lock); | |
1156 | ||
1157 | found = choose_mountpoint_rcu(m, root, path, &seq); | |
1158 | if (unlikely(!found)) { | |
1159 | if (!read_seqretry(&mount_lock, mseq)) | |
1160 | break; | |
1161 | } else { | |
1162 | if (likely(__legitimize_path(path, seq, mseq))) | |
1163 | break; | |
1164 | rcu_read_unlock(); | |
1165 | path_put(path); | |
1166 | rcu_read_lock(); | |
1167 | } | |
1168 | } | |
1169 | rcu_read_unlock(); | |
1170 | return found; | |
1171 | } | |
1172 | ||
b5c84bf6 | 1173 | /* |
9875cf80 DH |
1174 | * Perform an automount |
1175 | * - return -EISDIR to tell follow_managed() to stop and return the path we | |
1176 | * were called with. | |
1da177e4 | 1177 | */ |
1c9f5e06 | 1178 | static int follow_automount(struct path *path, int *count, unsigned lookup_flags) |
31e6b01f | 1179 | { |
25e195aa | 1180 | struct dentry *dentry = path->dentry; |
9875cf80 | 1181 | |
0ec26fd0 MS |
1182 | /* We don't want to mount if someone's just doing a stat - |
1183 | * unless they're stat'ing a directory and appended a '/' to | |
1184 | * the name. | |
1185 | * | |
1186 | * We do, however, want to mount if someone wants to open or | |
1187 | * create a file of any type under the mountpoint, wants to | |
1188 | * traverse through the mountpoint or wants to open the | |
1189 | * mounted directory. Also, autofs may mark negative dentries | |
1190 | * as being automount points. These will need the attentions | |
1191 | * of the daemon to instantiate them before they can be used. | |
9875cf80 | 1192 | */ |
1c9f5e06 | 1193 | if (!(lookup_flags & (LOOKUP_PARENT | LOOKUP_DIRECTORY | |
5d38f049 | 1194 | LOOKUP_OPEN | LOOKUP_CREATE | LOOKUP_AUTOMOUNT)) && |
25e195aa | 1195 | dentry->d_inode) |
5d38f049 | 1196 | return -EISDIR; |
0ec26fd0 | 1197 | |
1c9f5e06 | 1198 | if (count && (*count)++ >= MAXSYMLINKS) |
9875cf80 DH |
1199 | return -ELOOP; |
1200 | ||
25e195aa | 1201 | return finish_automount(dentry->d_op->d_automount(path), path); |
463ffb2e AV |
1202 | } |
1203 | ||
9875cf80 | 1204 | /* |
9deed3eb AV |
1205 | * mount traversal - out-of-line part. One note on ->d_flags accesses - |
1206 | * dentries are pinned but not locked here, so negative dentry can go | |
1207 | * positive right under us. Use of smp_load_acquire() provides a barrier | |
1208 | * sufficient for ->d_inode and ->d_flags consistency. | |
9875cf80 | 1209 | */ |
9deed3eb AV |
1210 | static int __traverse_mounts(struct path *path, unsigned flags, bool *jumped, |
1211 | int *count, unsigned lookup_flags) | |
1da177e4 | 1212 | { |
9deed3eb | 1213 | struct vfsmount *mnt = path->mnt; |
9875cf80 | 1214 | bool need_mntput = false; |
8aef1884 | 1215 | int ret = 0; |
9875cf80 | 1216 | |
9deed3eb | 1217 | while (flags & DCACHE_MANAGED_DENTRY) { |
cc53ce53 DH |
1218 | /* Allow the filesystem to manage the transit without i_mutex |
1219 | * being held. */ | |
d41efb52 | 1220 | if (flags & DCACHE_MANAGE_TRANSIT) { |
fb5f51c7 | 1221 | ret = path->dentry->d_op->d_manage(path, false); |
508c8772 | 1222 | flags = smp_load_acquire(&path->dentry->d_flags); |
cc53ce53 | 1223 | if (ret < 0) |
8aef1884 | 1224 | break; |
cc53ce53 DH |
1225 | } |
1226 | ||
9deed3eb | 1227 | if (flags & DCACHE_MOUNTED) { // something's mounted on it.. |
9875cf80 | 1228 | struct vfsmount *mounted = lookup_mnt(path); |
9deed3eb | 1229 | if (mounted) { // ... in our namespace |
9875cf80 DH |
1230 | dput(path->dentry); |
1231 | if (need_mntput) | |
1232 | mntput(path->mnt); | |
1233 | path->mnt = mounted; | |
1234 | path->dentry = dget(mounted->mnt_root); | |
9deed3eb AV |
1235 | // here we know it's positive |
1236 | flags = path->dentry->d_flags; | |
9875cf80 DH |
1237 | need_mntput = true; |
1238 | continue; | |
1239 | } | |
9875cf80 DH |
1240 | } |
1241 | ||
9deed3eb AV |
1242 | if (!(flags & DCACHE_NEED_AUTOMOUNT)) |
1243 | break; | |
9875cf80 | 1244 | |
9deed3eb AV |
1245 | // uncovered automount point |
1246 | ret = follow_automount(path, count, lookup_flags); | |
1247 | flags = smp_load_acquire(&path->dentry->d_flags); | |
1248 | if (ret < 0) | |
1249 | break; | |
1da177e4 | 1250 | } |
8aef1884 | 1251 | |
9deed3eb AV |
1252 | if (ret == -EISDIR) |
1253 | ret = 0; | |
1254 | // possible if you race with several mount --move | |
1255 | if (need_mntput && path->mnt == mnt) | |
1256 | mntput(path->mnt); | |
1257 | if (!ret && unlikely(d_flags_negative(flags))) | |
d41efb52 | 1258 | ret = -ENOENT; |
9deed3eb | 1259 | *jumped = need_mntput; |
8402752e | 1260 | return ret; |
1da177e4 LT |
1261 | } |
1262 | ||
9deed3eb AV |
1263 | static inline int traverse_mounts(struct path *path, bool *jumped, |
1264 | int *count, unsigned lookup_flags) | |
1265 | { | |
1266 | unsigned flags = smp_load_acquire(&path->dentry->d_flags); | |
1267 | ||
1268 | /* fastpath */ | |
1269 | if (likely(!(flags & DCACHE_MANAGED_DENTRY))) { | |
1270 | *jumped = false; | |
1271 | if (unlikely(d_flags_negative(flags))) | |
1272 | return -ENOENT; | |
1273 | return 0; | |
1274 | } | |
1275 | return __traverse_mounts(path, flags, jumped, count, lookup_flags); | |
1276 | } | |
1277 | ||
cc53ce53 | 1278 | int follow_down_one(struct path *path) |
1da177e4 LT |
1279 | { |
1280 | struct vfsmount *mounted; | |
1281 | ||
1c755af4 | 1282 | mounted = lookup_mnt(path); |
1da177e4 | 1283 | if (mounted) { |
9393bd07 AV |
1284 | dput(path->dentry); |
1285 | mntput(path->mnt); | |
1286 | path->mnt = mounted; | |
1287 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
1288 | return 1; |
1289 | } | |
1290 | return 0; | |
1291 | } | |
4d359507 | 1292 | EXPORT_SYMBOL(follow_down_one); |
1da177e4 | 1293 | |
9deed3eb AV |
1294 | /* |
1295 | * Follow down to the covering mount currently visible to userspace. At each | |
1296 | * point, the filesystem owning that dentry may be queried as to whether the | |
1297 | * caller is permitted to proceed or not. | |
1298 | */ | |
1299 | int follow_down(struct path *path) | |
1300 | { | |
1301 | struct vfsmount *mnt = path->mnt; | |
1302 | bool jumped; | |
1303 | int ret = traverse_mounts(path, &jumped, NULL, 0); | |
1304 | ||
1305 | if (path->mnt != mnt) | |
1306 | mntput(mnt); | |
1307 | return ret; | |
1308 | } | |
1309 | EXPORT_SYMBOL(follow_down); | |
1310 | ||
9875cf80 | 1311 | /* |
287548e4 AV |
1312 | * Try to skip to top of mountpoint pile in rcuwalk mode. Fail if |
1313 | * we meet a managed dentry that would need blocking. | |
9875cf80 DH |
1314 | */ |
1315 | static bool __follow_mount_rcu(struct nameidata *nd, struct path *path, | |
254cf582 | 1316 | struct inode **inode, unsigned *seqp) |
9875cf80 | 1317 | { |
ea936aeb AV |
1318 | struct dentry *dentry = path->dentry; |
1319 | unsigned int flags = dentry->d_flags; | |
1320 | ||
1321 | if (likely(!(flags & DCACHE_MANAGED_DENTRY))) | |
1322 | return true; | |
1323 | ||
1324 | if (unlikely(nd->flags & LOOKUP_NO_XDEV)) | |
1325 | return false; | |
1326 | ||
62a7375e | 1327 | for (;;) { |
62a7375e IK |
1328 | /* |
1329 | * Don't forget we might have a non-mountpoint managed dentry | |
1330 | * that wants to block transit. | |
1331 | */ | |
ea936aeb AV |
1332 | if (unlikely(flags & DCACHE_MANAGE_TRANSIT)) { |
1333 | int res = dentry->d_op->d_manage(path, true); | |
1334 | if (res) | |
1335 | return res == -EISDIR; | |
1336 | flags = dentry->d_flags; | |
b8faf035 | 1337 | } |
62a7375e | 1338 | |
ea936aeb AV |
1339 | if (flags & DCACHE_MOUNTED) { |
1340 | struct mount *mounted = __lookup_mnt(path->mnt, dentry); | |
1341 | if (mounted) { | |
1342 | path->mnt = &mounted->mnt; | |
1343 | dentry = path->dentry = mounted->mnt.mnt_root; | |
1344 | nd->flags |= LOOKUP_JUMPED; | |
1345 | *seqp = read_seqcount_begin(&dentry->d_seq); | |
1346 | *inode = dentry->d_inode; | |
1347 | /* | |
1348 | * We don't need to re-check ->d_seq after this | |
1349 | * ->d_inode read - there will be an RCU delay | |
1350 | * between mount hash removal and ->mnt_root | |
1351 | * becoming unpinned. | |
1352 | */ | |
1353 | flags = dentry->d_flags; | |
1354 | continue; | |
1355 | } | |
1356 | if (read_seqretry(&mount_lock, nd->m_seq)) | |
1357 | return false; | |
1358 | } | |
1359 | return !(flags & DCACHE_NEED_AUTOMOUNT); | |
9875cf80 | 1360 | } |
287548e4 AV |
1361 | } |
1362 | ||
db3c9ade AV |
1363 | static inline int handle_mounts(struct nameidata *nd, struct dentry *dentry, |
1364 | struct path *path, struct inode **inode, | |
1365 | unsigned int *seqp) | |
bd7c4b50 | 1366 | { |
9deed3eb | 1367 | bool jumped; |
db3c9ade | 1368 | int ret; |
bd7c4b50 | 1369 | |
db3c9ade AV |
1370 | path->mnt = nd->path.mnt; |
1371 | path->dentry = dentry; | |
c153007b AV |
1372 | if (nd->flags & LOOKUP_RCU) { |
1373 | unsigned int seq = *seqp; | |
1374 | if (unlikely(!*inode)) | |
1375 | return -ENOENT; | |
1376 | if (likely(__follow_mount_rcu(nd, path, inode, seqp))) | |
9deed3eb | 1377 | return 0; |
c153007b AV |
1378 | if (unlazy_child(nd, dentry, seq)) |
1379 | return -ECHILD; | |
1380 | // *path might've been clobbered by __follow_mount_rcu() | |
1381 | path->mnt = nd->path.mnt; | |
1382 | path->dentry = dentry; | |
1383 | } | |
9deed3eb AV |
1384 | ret = traverse_mounts(path, &jumped, &nd->total_link_count, nd->flags); |
1385 | if (jumped) { | |
1386 | if (unlikely(nd->flags & LOOKUP_NO_XDEV)) | |
1387 | ret = -EXDEV; | |
1388 | else | |
1389 | nd->flags |= LOOKUP_JUMPED; | |
1390 | } | |
1391 | if (unlikely(ret)) { | |
1392 | dput(path->dentry); | |
1393 | if (path->mnt != nd->path.mnt) | |
1394 | mntput(path->mnt); | |
1395 | } else { | |
bd7c4b50 AV |
1396 | *inode = d_backing_inode(path->dentry); |
1397 | *seqp = 0; /* out of RCU mode, so the value doesn't matter */ | |
1398 | } | |
1399 | return ret; | |
1400 | } | |
1401 | ||
baa03890 | 1402 | /* |
f4fdace9 OD |
1403 | * This looks up the name in dcache and possibly revalidates the found dentry. |
1404 | * NULL is returned if the dentry does not exist in the cache. | |
baa03890 | 1405 | */ |
e3c13928 AV |
1406 | static struct dentry *lookup_dcache(const struct qstr *name, |
1407 | struct dentry *dir, | |
6c51e513 | 1408 | unsigned int flags) |
baa03890 | 1409 | { |
a89f8337 | 1410 | struct dentry *dentry = d_lookup(dir, name); |
bad61189 | 1411 | if (dentry) { |
a89f8337 AV |
1412 | int error = d_revalidate(dentry, flags); |
1413 | if (unlikely(error <= 0)) { | |
1414 | if (!error) | |
1415 | d_invalidate(dentry); | |
1416 | dput(dentry); | |
1417 | return ERR_PTR(error); | |
bad61189 MS |
1418 | } |
1419 | } | |
baa03890 NP |
1420 | return dentry; |
1421 | } | |
1422 | ||
44396f4b | 1423 | /* |
a03ece5f AV |
1424 | * Parent directory has inode locked exclusive. This is one |
1425 | * and only case when ->lookup() gets called on non in-lookup | |
1426 | * dentries - as the matter of fact, this only gets called | |
1427 | * when directory is guaranteed to have no in-lookup children | |
1428 | * at all. | |
44396f4b | 1429 | */ |
e3c13928 | 1430 | static struct dentry *__lookup_hash(const struct qstr *name, |
72bd866a | 1431 | struct dentry *base, unsigned int flags) |
a3255546 | 1432 | { |
6c51e513 | 1433 | struct dentry *dentry = lookup_dcache(name, base, flags); |
a03ece5f AV |
1434 | struct dentry *old; |
1435 | struct inode *dir = base->d_inode; | |
a3255546 | 1436 | |
6c51e513 | 1437 | if (dentry) |
bad61189 | 1438 | return dentry; |
a3255546 | 1439 | |
a03ece5f AV |
1440 | /* Don't create child dentry for a dead directory. */ |
1441 | if (unlikely(IS_DEADDIR(dir))) | |
1442 | return ERR_PTR(-ENOENT); | |
1443 | ||
6c51e513 AV |
1444 | dentry = d_alloc(base, name); |
1445 | if (unlikely(!dentry)) | |
1446 | return ERR_PTR(-ENOMEM); | |
1447 | ||
a03ece5f AV |
1448 | old = dir->i_op->lookup(dir, dentry, flags); |
1449 | if (unlikely(old)) { | |
1450 | dput(dentry); | |
1451 | dentry = old; | |
1452 | } | |
1453 | return dentry; | |
a3255546 AV |
1454 | } |
1455 | ||
20e34357 AV |
1456 | static struct dentry *lookup_fast(struct nameidata *nd, |
1457 | struct inode **inode, | |
1458 | unsigned *seqp) | |
1da177e4 | 1459 | { |
31e6b01f | 1460 | struct dentry *dentry, *parent = nd->path.dentry; |
5a18fff2 | 1461 | int status = 1; |
9875cf80 | 1462 | |
b04f784e NP |
1463 | /* |
1464 | * Rename seqlock is not required here because in the off chance | |
5d0f49c1 AV |
1465 | * of a false negative due to a concurrent rename, the caller is |
1466 | * going to fall back to non-racy lookup. | |
b04f784e | 1467 | */ |
31e6b01f NP |
1468 | if (nd->flags & LOOKUP_RCU) { |
1469 | unsigned seq; | |
da53be12 | 1470 | dentry = __d_lookup_rcu(parent, &nd->last, &seq); |
5d0f49c1 | 1471 | if (unlikely(!dentry)) { |
4675ac39 | 1472 | if (unlazy_walk(nd)) |
20e34357 AV |
1473 | return ERR_PTR(-ECHILD); |
1474 | return NULL; | |
5d0f49c1 | 1475 | } |
5a18fff2 | 1476 | |
12f8ad4b LT |
1477 | /* |
1478 | * This sequence count validates that the inode matches | |
1479 | * the dentry name information from lookup. | |
1480 | */ | |
63afdfc7 | 1481 | *inode = d_backing_inode(dentry); |
5d0f49c1 | 1482 | if (unlikely(read_seqcount_retry(&dentry->d_seq, seq))) |
20e34357 | 1483 | return ERR_PTR(-ECHILD); |
12f8ad4b LT |
1484 | |
1485 | /* | |
1486 | * This sequence count validates that the parent had no | |
1487 | * changes while we did the lookup of the dentry above. | |
1488 | * | |
1489 | * The memory barrier in read_seqcount_begin of child is | |
1490 | * enough, we can use __read_seqcount_retry here. | |
1491 | */ | |
5d0f49c1 | 1492 | if (unlikely(__read_seqcount_retry(&parent->d_seq, nd->seq))) |
20e34357 | 1493 | return ERR_PTR(-ECHILD); |
5a18fff2 | 1494 | |
254cf582 | 1495 | *seqp = seq; |
a89f8337 | 1496 | status = d_revalidate(dentry, nd->flags); |
c153007b | 1497 | if (likely(status > 0)) |
20e34357 | 1498 | return dentry; |
4675ac39 | 1499 | if (unlazy_child(nd, dentry, seq)) |
20e34357 | 1500 | return ERR_PTR(-ECHILD); |
209a7fb2 AV |
1501 | if (unlikely(status == -ECHILD)) |
1502 | /* we'd been told to redo it in non-rcu mode */ | |
1503 | status = d_revalidate(dentry, nd->flags); | |
5a18fff2 | 1504 | } else { |
e97cdc87 | 1505 | dentry = __d_lookup(parent, &nd->last); |
5d0f49c1 | 1506 | if (unlikely(!dentry)) |
20e34357 | 1507 | return NULL; |
a89f8337 | 1508 | status = d_revalidate(dentry, nd->flags); |
9875cf80 | 1509 | } |
5a18fff2 | 1510 | if (unlikely(status <= 0)) { |
e9742b53 | 1511 | if (!status) |
5d0f49c1 | 1512 | d_invalidate(dentry); |
5542aa2f | 1513 | dput(dentry); |
20e34357 | 1514 | return ERR_PTR(status); |
24643087 | 1515 | } |
20e34357 | 1516 | return dentry; |
697f514d MS |
1517 | } |
1518 | ||
1519 | /* Fast lookup failed, do it the slow way */ | |
88d8331a AV |
1520 | static struct dentry *__lookup_slow(const struct qstr *name, |
1521 | struct dentry *dir, | |
1522 | unsigned int flags) | |
697f514d | 1523 | { |
88d8331a | 1524 | struct dentry *dentry, *old; |
1936386e | 1525 | struct inode *inode = dir->d_inode; |
d9171b93 | 1526 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
1936386e | 1527 | |
1936386e | 1528 | /* Don't go there if it's already dead */ |
94bdd655 | 1529 | if (unlikely(IS_DEADDIR(inode))) |
88d8331a | 1530 | return ERR_PTR(-ENOENT); |
94bdd655 | 1531 | again: |
d9171b93 | 1532 | dentry = d_alloc_parallel(dir, name, &wq); |
94bdd655 | 1533 | if (IS_ERR(dentry)) |
88d8331a | 1534 | return dentry; |
94bdd655 | 1535 | if (unlikely(!d_in_lookup(dentry))) { |
c64cd6e3 AV |
1536 | int error = d_revalidate(dentry, flags); |
1537 | if (unlikely(error <= 0)) { | |
1538 | if (!error) { | |
1539 | d_invalidate(dentry); | |
949a852e | 1540 | dput(dentry); |
c64cd6e3 | 1541 | goto again; |
949a852e | 1542 | } |
c64cd6e3 AV |
1543 | dput(dentry); |
1544 | dentry = ERR_PTR(error); | |
949a852e | 1545 | } |
94bdd655 AV |
1546 | } else { |
1547 | old = inode->i_op->lookup(inode, dentry, flags); | |
1548 | d_lookup_done(dentry); | |
1549 | if (unlikely(old)) { | |
1550 | dput(dentry); | |
1551 | dentry = old; | |
949a852e AV |
1552 | } |
1553 | } | |
e3c13928 | 1554 | return dentry; |
1da177e4 LT |
1555 | } |
1556 | ||
88d8331a AV |
1557 | static struct dentry *lookup_slow(const struct qstr *name, |
1558 | struct dentry *dir, | |
1559 | unsigned int flags) | |
1560 | { | |
1561 | struct inode *inode = dir->d_inode; | |
1562 | struct dentry *res; | |
1563 | inode_lock_shared(inode); | |
1564 | res = __lookup_slow(name, dir, flags); | |
1565 | inode_unlock_shared(inode); | |
1566 | return res; | |
1567 | } | |
1568 | ||
52094c8a AV |
1569 | static inline int may_lookup(struct nameidata *nd) |
1570 | { | |
1571 | if (nd->flags & LOOKUP_RCU) { | |
4ad5abb3 | 1572 | int err = inode_permission(nd->inode, MAY_EXEC|MAY_NOT_BLOCK); |
52094c8a AV |
1573 | if (err != -ECHILD) |
1574 | return err; | |
4675ac39 | 1575 | if (unlazy_walk(nd)) |
52094c8a AV |
1576 | return -ECHILD; |
1577 | } | |
4ad5abb3 | 1578 | return inode_permission(nd->inode, MAY_EXEC); |
52094c8a AV |
1579 | } |
1580 | ||
49055906 AV |
1581 | static int reserve_stack(struct nameidata *nd, struct path *link, unsigned seq) |
1582 | { | |
49055906 AV |
1583 | if (unlikely(nd->total_link_count++ >= MAXSYMLINKS)) |
1584 | return -ELOOP; | |
4542576b AV |
1585 | |
1586 | if (likely(nd->depth != EMBEDDED_LEVELS)) | |
1587 | return 0; | |
1588 | if (likely(nd->stack != nd->internal)) | |
1589 | return 0; | |
60ef60c7 | 1590 | if (likely(nd_alloc_stack(nd))) |
49055906 | 1591 | return 0; |
60ef60c7 AV |
1592 | |
1593 | if (nd->flags & LOOKUP_RCU) { | |
1594 | // we need to grab link before we do unlazy. And we can't skip | |
1595 | // unlazy even if we fail to grab the link - cleanup needs it | |
49055906 | 1596 | bool grabbed_link = legitimize_path(nd, link, seq); |
60ef60c7 AV |
1597 | |
1598 | if (unlazy_walk(nd) != 0 || !grabbed_link) | |
1599 | return -ECHILD; | |
1600 | ||
1601 | if (nd_alloc_stack(nd)) | |
1602 | return 0; | |
49055906 | 1603 | } |
60ef60c7 | 1604 | return -ENOMEM; |
49055906 AV |
1605 | } |
1606 | ||
b1a81972 AV |
1607 | enum {WALK_TRAILING = 1, WALK_MORE = 2, WALK_NOFOLLOW = 4}; |
1608 | ||
06708adb | 1609 | static const char *pick_link(struct nameidata *nd, struct path *link, |
b1a81972 | 1610 | struct inode *inode, unsigned seq, int flags) |
d63ff28f | 1611 | { |
1cf2665b | 1612 | struct saved *last; |
ad6cc4c3 | 1613 | const char *res; |
49055906 | 1614 | int error = reserve_stack(nd, link, seq); |
ad6cc4c3 | 1615 | |
626de996 | 1616 | if (unlikely(error)) { |
49055906 | 1617 | if (!(nd->flags & LOOKUP_RCU)) |
bc40aee0 | 1618 | path_put(link); |
49055906 | 1619 | return ERR_PTR(error); |
626de996 | 1620 | } |
ab104923 | 1621 | last = nd->stack + nd->depth++; |
1cf2665b | 1622 | last->link = *link; |
fceef393 | 1623 | clear_delayed_call(&last->done); |
0450b2d1 | 1624 | last->seq = seq; |
ad6cc4c3 | 1625 | |
b1a81972 | 1626 | if (flags & WALK_TRAILING) { |
ad6cc4c3 AV |
1627 | error = may_follow_link(nd, inode); |
1628 | if (unlikely(error)) | |
1629 | return ERR_PTR(error); | |
1630 | } | |
1631 | ||
dab741e0 MN |
1632 | if (unlikely(nd->flags & LOOKUP_NO_SYMLINKS) || |
1633 | unlikely(link->mnt->mnt_flags & MNT_NOSYMFOLLOW)) | |
ad6cc4c3 AV |
1634 | return ERR_PTR(-ELOOP); |
1635 | ||
1636 | if (!(nd->flags & LOOKUP_RCU)) { | |
1637 | touch_atime(&last->link); | |
1638 | cond_resched(); | |
1639 | } else if (atime_needs_update(&last->link, inode)) { | |
1640 | if (unlikely(unlazy_walk(nd))) | |
1641 | return ERR_PTR(-ECHILD); | |
1642 | touch_atime(&last->link); | |
1643 | } | |
1644 | ||
1645 | error = security_inode_follow_link(link->dentry, inode, | |
1646 | nd->flags & LOOKUP_RCU); | |
1647 | if (unlikely(error)) | |
1648 | return ERR_PTR(error); | |
1649 | ||
ad6cc4c3 AV |
1650 | res = READ_ONCE(inode->i_link); |
1651 | if (!res) { | |
1652 | const char * (*get)(struct dentry *, struct inode *, | |
1653 | struct delayed_call *); | |
1654 | get = inode->i_op->get_link; | |
1655 | if (nd->flags & LOOKUP_RCU) { | |
1656 | res = get(NULL, inode, &last->done); | |
1657 | if (res == ERR_PTR(-ECHILD)) { | |
1658 | if (unlikely(unlazy_walk(nd))) | |
1659 | return ERR_PTR(-ECHILD); | |
1660 | res = get(link->dentry, inode, &last->done); | |
1661 | } | |
1662 | } else { | |
1663 | res = get(link->dentry, inode, &last->done); | |
1664 | } | |
1665 | if (!res) | |
1666 | goto all_done; | |
1667 | if (IS_ERR(res)) | |
1668 | return res; | |
1669 | } | |
1670 | if (*res == '/') { | |
1671 | error = nd_jump_root(nd); | |
1672 | if (unlikely(error)) | |
1673 | return ERR_PTR(error); | |
1674 | while (unlikely(*++res == '/')) | |
1675 | ; | |
1676 | } | |
1677 | if (*res) | |
1678 | return res; | |
1679 | all_done: // pure jump | |
1680 | put_link(nd); | |
1681 | return NULL; | |
d63ff28f AV |
1682 | } |
1683 | ||
3ddcd056 LT |
1684 | /* |
1685 | * Do we need to follow links? We _really_ want to be able | |
1686 | * to do this check without having to look at inode->i_op, | |
1687 | * so we keep a cache of "no, this doesn't need follow_link" | |
1688 | * for the common case. | |
1689 | */ | |
b0417d2c | 1690 | static const char *step_into(struct nameidata *nd, int flags, |
cbae4d12 | 1691 | struct dentry *dentry, struct inode *inode, unsigned seq) |
3ddcd056 | 1692 | { |
cbae4d12 AV |
1693 | struct path path; |
1694 | int err = handle_mounts(nd, dentry, &path, &inode, &seq); | |
1695 | ||
1696 | if (err < 0) | |
b0417d2c | 1697 | return ERR_PTR(err); |
cbae4d12 | 1698 | if (likely(!d_is_symlink(path.dentry)) || |
8c4efe22 | 1699 | ((flags & WALK_TRAILING) && !(nd->flags & LOOKUP_FOLLOW)) || |
aca2903e | 1700 | (flags & WALK_NOFOLLOW)) { |
8f64fb1c | 1701 | /* not a symlink or should not follow */ |
c99687a0 AV |
1702 | if (!(nd->flags & LOOKUP_RCU)) { |
1703 | dput(nd->path.dentry); | |
1704 | if (nd->path.mnt != path.mnt) | |
1705 | mntput(nd->path.mnt); | |
1706 | } | |
1707 | nd->path = path; | |
8f64fb1c AV |
1708 | nd->inode = inode; |
1709 | nd->seq = seq; | |
b0417d2c | 1710 | return NULL; |
8f64fb1c | 1711 | } |
a7f77542 | 1712 | if (nd->flags & LOOKUP_RCU) { |
84f0cd9e | 1713 | /* make sure that d_is_symlink above matches inode */ |
cbae4d12 | 1714 | if (read_seqcount_retry(&path.dentry->d_seq, seq)) |
b0417d2c | 1715 | return ERR_PTR(-ECHILD); |
84f0cd9e AV |
1716 | } else { |
1717 | if (path.mnt == nd->path.mnt) | |
1718 | mntget(path.mnt); | |
a7f77542 | 1719 | } |
b1a81972 | 1720 | return pick_link(nd, &path, inode, seq, flags); |
3ddcd056 LT |
1721 | } |
1722 | ||
c2df1968 AV |
1723 | static struct dentry *follow_dotdot_rcu(struct nameidata *nd, |
1724 | struct inode **inodep, | |
1725 | unsigned *seqp) | |
957dd41d | 1726 | { |
12487f30 | 1727 | struct dentry *parent, *old; |
957dd41d | 1728 | |
12487f30 AV |
1729 | if (path_equal(&nd->path, &nd->root)) |
1730 | goto in_root; | |
1731 | if (unlikely(nd->path.dentry == nd->path.mnt->mnt_root)) { | |
7ef482fa | 1732 | struct path path; |
efe772d6 | 1733 | unsigned seq; |
7ef482fa AV |
1734 | if (!choose_mountpoint_rcu(real_mount(nd->path.mnt), |
1735 | &nd->root, &path, &seq)) | |
1736 | goto in_root; | |
efe772d6 AV |
1737 | if (unlikely(nd->flags & LOOKUP_NO_XDEV)) |
1738 | return ERR_PTR(-ECHILD); | |
1739 | nd->path = path; | |
1740 | nd->inode = path.dentry->d_inode; | |
1741 | nd->seq = seq; | |
1742 | if (unlikely(read_seqretry(&mount_lock, nd->m_seq))) | |
1743 | return ERR_PTR(-ECHILD); | |
1744 | /* we know that mountpoint was pinned */ | |
957dd41d | 1745 | } |
12487f30 AV |
1746 | old = nd->path.dentry; |
1747 | parent = old->d_parent; | |
1748 | *inodep = parent->d_inode; | |
1749 | *seqp = read_seqcount_begin(&parent->d_seq); | |
1750 | if (unlikely(read_seqcount_retry(&old->d_seq, nd->seq))) | |
1751 | return ERR_PTR(-ECHILD); | |
1752 | if (unlikely(!path_connected(nd->path.mnt, parent))) | |
1753 | return ERR_PTR(-ECHILD); | |
1754 | return parent; | |
1755 | in_root: | |
efe772d6 AV |
1756 | if (unlikely(read_seqretry(&mount_lock, nd->m_seq))) |
1757 | return ERR_PTR(-ECHILD); | |
c2df1968 AV |
1758 | if (unlikely(nd->flags & LOOKUP_BENEATH)) |
1759 | return ERR_PTR(-ECHILD); | |
1760 | return NULL; | |
957dd41d AV |
1761 | } |
1762 | ||
c2df1968 AV |
1763 | static struct dentry *follow_dotdot(struct nameidata *nd, |
1764 | struct inode **inodep, | |
1765 | unsigned *seqp) | |
957dd41d | 1766 | { |
12487f30 AV |
1767 | struct dentry *parent; |
1768 | ||
1769 | if (path_equal(&nd->path, &nd->root)) | |
1770 | goto in_root; | |
1771 | if (unlikely(nd->path.dentry == nd->path.mnt->mnt_root)) { | |
2aa38470 AV |
1772 | struct path path; |
1773 | ||
1774 | if (!choose_mountpoint(real_mount(nd->path.mnt), | |
1775 | &nd->root, &path)) | |
1776 | goto in_root; | |
165200d6 AV |
1777 | path_put(&nd->path); |
1778 | nd->path = path; | |
2aa38470 | 1779 | nd->inode = path.dentry->d_inode; |
165200d6 AV |
1780 | if (unlikely(nd->flags & LOOKUP_NO_XDEV)) |
1781 | return ERR_PTR(-EXDEV); | |
957dd41d | 1782 | } |
12487f30 AV |
1783 | /* rare case of legitimate dget_parent()... */ |
1784 | parent = dget_parent(nd->path.dentry); | |
1785 | if (unlikely(!path_connected(nd->path.mnt, parent))) { | |
1786 | dput(parent); | |
1787 | return ERR_PTR(-ENOENT); | |
1788 | } | |
1789 | *seqp = 0; | |
1790 | *inodep = parent->d_inode; | |
1791 | return parent; | |
1792 | ||
1793 | in_root: | |
c2df1968 AV |
1794 | if (unlikely(nd->flags & LOOKUP_BENEATH)) |
1795 | return ERR_PTR(-EXDEV); | |
1796 | dget(nd->path.dentry); | |
1797 | return NULL; | |
957dd41d AV |
1798 | } |
1799 | ||
7521f22b | 1800 | static const char *handle_dots(struct nameidata *nd, int type) |
957dd41d AV |
1801 | { |
1802 | if (type == LAST_DOTDOT) { | |
7521f22b | 1803 | const char *error = NULL; |
c2df1968 AV |
1804 | struct dentry *parent; |
1805 | struct inode *inode; | |
1806 | unsigned seq; | |
957dd41d AV |
1807 | |
1808 | if (!nd->root.mnt) { | |
7521f22b | 1809 | error = ERR_PTR(set_root(nd)); |
957dd41d AV |
1810 | if (error) |
1811 | return error; | |
1812 | } | |
1813 | if (nd->flags & LOOKUP_RCU) | |
c2df1968 | 1814 | parent = follow_dotdot_rcu(nd, &inode, &seq); |
957dd41d | 1815 | else |
c2df1968 AV |
1816 | parent = follow_dotdot(nd, &inode, &seq); |
1817 | if (IS_ERR(parent)) | |
1818 | return ERR_CAST(parent); | |
1819 | if (unlikely(!parent)) | |
1820 | error = step_into(nd, WALK_NOFOLLOW, | |
1821 | nd->path.dentry, nd->inode, nd->seq); | |
1822 | else | |
1823 | error = step_into(nd, WALK_NOFOLLOW, | |
1824 | parent, inode, seq); | |
1825 | if (unlikely(error)) | |
957dd41d AV |
1826 | return error; |
1827 | ||
1828 | if (unlikely(nd->flags & LOOKUP_IS_SCOPED)) { | |
1829 | /* | |
1830 | * If there was a racing rename or mount along our | |
1831 | * path, then we can't be sure that ".." hasn't jumped | |
1832 | * above nd->root (and so userspace should retry or use | |
1833 | * some fallback). | |
1834 | */ | |
1835 | smp_rmb(); | |
1836 | if (unlikely(__read_seqcount_retry(&mount_lock.seqcount, nd->m_seq))) | |
7521f22b | 1837 | return ERR_PTR(-EAGAIN); |
957dd41d | 1838 | if (unlikely(__read_seqcount_retry(&rename_lock.seqcount, nd->r_seq))) |
7521f22b | 1839 | return ERR_PTR(-EAGAIN); |
957dd41d AV |
1840 | } |
1841 | } | |
7521f22b | 1842 | return NULL; |
957dd41d AV |
1843 | } |
1844 | ||
92d27016 | 1845 | static const char *walk_component(struct nameidata *nd, int flags) |
ce57dfc1 | 1846 | { |
db3c9ade | 1847 | struct dentry *dentry; |
ce57dfc1 | 1848 | struct inode *inode; |
254cf582 | 1849 | unsigned seq; |
ce57dfc1 AV |
1850 | /* |
1851 | * "." and ".." are special - ".." especially so because it has | |
1852 | * to be able to know about the current root directory and | |
1853 | * parent relationships. | |
1854 | */ | |
4693a547 | 1855 | if (unlikely(nd->last_type != LAST_NORM)) { |
1c4ff1a8 | 1856 | if (!(flags & WALK_MORE) && nd->depth) |
4693a547 | 1857 | put_link(nd); |
7521f22b | 1858 | return handle_dots(nd, nd->last_type); |
4693a547 | 1859 | } |
20e34357 AV |
1860 | dentry = lookup_fast(nd, &inode, &seq); |
1861 | if (IS_ERR(dentry)) | |
92d27016 | 1862 | return ERR_CAST(dentry); |
20e34357 | 1863 | if (unlikely(!dentry)) { |
db3c9ade AV |
1864 | dentry = lookup_slow(&nd->last, nd->path.dentry, nd->flags); |
1865 | if (IS_ERR(dentry)) | |
92d27016 | 1866 | return ERR_CAST(dentry); |
ce57dfc1 | 1867 | } |
56676ec3 AV |
1868 | if (!(flags & WALK_MORE) && nd->depth) |
1869 | put_link(nd); | |
b0417d2c | 1870 | return step_into(nd, flags, dentry, inode, seq); |
ce57dfc1 AV |
1871 | } |
1872 | ||
bfcfaa77 LT |
1873 | /* |
1874 | * We can do the critical dentry name comparison and hashing | |
1875 | * operations one word at a time, but we are limited to: | |
1876 | * | |
1877 | * - Architectures with fast unaligned word accesses. We could | |
1878 | * do a "get_unaligned()" if this helps and is sufficiently | |
1879 | * fast. | |
1880 | * | |
bfcfaa77 LT |
1881 | * - non-CONFIG_DEBUG_PAGEALLOC configurations (so that we |
1882 | * do not trap on the (extremely unlikely) case of a page | |
1883 | * crossing operation. | |
1884 | * | |
1885 | * - Furthermore, we need an efficient 64-bit compile for the | |
1886 | * 64-bit case in order to generate the "number of bytes in | |
1887 | * the final mask". Again, that could be replaced with a | |
1888 | * efficient population count instruction or similar. | |
1889 | */ | |
1890 | #ifdef CONFIG_DCACHE_WORD_ACCESS | |
1891 | ||
f68e556e | 1892 | #include <asm/word-at-a-time.h> |
bfcfaa77 | 1893 | |
468a9428 | 1894 | #ifdef HASH_MIX |
bfcfaa77 | 1895 | |
468a9428 | 1896 | /* Architecture provides HASH_MIX and fold_hash() in <asm/hash.h> */ |
bfcfaa77 | 1897 | |
468a9428 | 1898 | #elif defined(CONFIG_64BIT) |
0fed3ac8 | 1899 | /* |
2a18da7a GS |
1900 | * Register pressure in the mixing function is an issue, particularly |
1901 | * on 32-bit x86, but almost any function requires one state value and | |
1902 | * one temporary. Instead, use a function designed for two state values | |
1903 | * and no temporaries. | |
1904 | * | |
1905 | * This function cannot create a collision in only two iterations, so | |
1906 | * we have two iterations to achieve avalanche. In those two iterations, | |
1907 | * we have six layers of mixing, which is enough to spread one bit's | |
1908 | * influence out to 2^6 = 64 state bits. | |
1909 | * | |
1910 | * Rotate constants are scored by considering either 64 one-bit input | |
1911 | * deltas or 64*63/2 = 2016 two-bit input deltas, and finding the | |
1912 | * probability of that delta causing a change to each of the 128 output | |
1913 | * bits, using a sample of random initial states. | |
1914 | * | |
1915 | * The Shannon entropy of the computed probabilities is then summed | |
1916 | * to produce a score. Ideally, any input change has a 50% chance of | |
1917 | * toggling any given output bit. | |
1918 | * | |
1919 | * Mixing scores (in bits) for (12,45): | |
1920 | * Input delta: 1-bit 2-bit | |
1921 | * 1 round: 713.3 42542.6 | |
1922 | * 2 rounds: 2753.7 140389.8 | |
1923 | * 3 rounds: 5954.1 233458.2 | |
1924 | * 4 rounds: 7862.6 256672.2 | |
1925 | * Perfect: 8192 258048 | |
1926 | * (64*128) (64*63/2 * 128) | |
0fed3ac8 | 1927 | */ |
2a18da7a GS |
1928 | #define HASH_MIX(x, y, a) \ |
1929 | ( x ^= (a), \ | |
1930 | y ^= x, x = rol64(x,12),\ | |
1931 | x += y, y = rol64(y,45),\ | |
1932 | y *= 9 ) | |
bfcfaa77 | 1933 | |
0fed3ac8 | 1934 | /* |
2a18da7a GS |
1935 | * Fold two longs into one 32-bit hash value. This must be fast, but |
1936 | * latency isn't quite as critical, as there is a fair bit of additional | |
1937 | * work done before the hash value is used. | |
0fed3ac8 | 1938 | */ |
2a18da7a | 1939 | static inline unsigned int fold_hash(unsigned long x, unsigned long y) |
0fed3ac8 | 1940 | { |
2a18da7a GS |
1941 | y ^= x * GOLDEN_RATIO_64; |
1942 | y *= GOLDEN_RATIO_64; | |
1943 | return y >> 32; | |
0fed3ac8 GS |
1944 | } |
1945 | ||
bfcfaa77 LT |
1946 | #else /* 32-bit case */ |
1947 | ||
2a18da7a GS |
1948 | /* |
1949 | * Mixing scores (in bits) for (7,20): | |
1950 | * Input delta: 1-bit 2-bit | |
1951 | * 1 round: 330.3 9201.6 | |
1952 | * 2 rounds: 1246.4 25475.4 | |
1953 | * 3 rounds: 1907.1 31295.1 | |
1954 | * 4 rounds: 2042.3 31718.6 | |
1955 | * Perfect: 2048 31744 | |
1956 | * (32*64) (32*31/2 * 64) | |
1957 | */ | |
1958 | #define HASH_MIX(x, y, a) \ | |
1959 | ( x ^= (a), \ | |
1960 | y ^= x, x = rol32(x, 7),\ | |
1961 | x += y, y = rol32(y,20),\ | |
1962 | y *= 9 ) | |
bfcfaa77 | 1963 | |
2a18da7a | 1964 | static inline unsigned int fold_hash(unsigned long x, unsigned long y) |
0fed3ac8 | 1965 | { |
2a18da7a GS |
1966 | /* Use arch-optimized multiply if one exists */ |
1967 | return __hash_32(y ^ __hash_32(x)); | |
0fed3ac8 GS |
1968 | } |
1969 | ||
bfcfaa77 LT |
1970 | #endif |
1971 | ||
2a18da7a GS |
1972 | /* |
1973 | * Return the hash of a string of known length. This is carfully | |
1974 | * designed to match hash_name(), which is the more critical function. | |
1975 | * In particular, we must end by hashing a final word containing 0..7 | |
1976 | * payload bytes, to match the way that hash_name() iterates until it | |
1977 | * finds the delimiter after the name. | |
1978 | */ | |
8387ff25 | 1979 | unsigned int full_name_hash(const void *salt, const char *name, unsigned int len) |
bfcfaa77 | 1980 | { |
8387ff25 | 1981 | unsigned long a, x = 0, y = (unsigned long)salt; |
bfcfaa77 LT |
1982 | |
1983 | for (;;) { | |
fcfd2fbf GS |
1984 | if (!len) |
1985 | goto done; | |
e419b4cc | 1986 | a = load_unaligned_zeropad(name); |
bfcfaa77 LT |
1987 | if (len < sizeof(unsigned long)) |
1988 | break; | |
2a18da7a | 1989 | HASH_MIX(x, y, a); |
bfcfaa77 LT |
1990 | name += sizeof(unsigned long); |
1991 | len -= sizeof(unsigned long); | |
bfcfaa77 | 1992 | } |
2a18da7a | 1993 | x ^= a & bytemask_from_count(len); |
bfcfaa77 | 1994 | done: |
2a18da7a | 1995 | return fold_hash(x, y); |
bfcfaa77 LT |
1996 | } |
1997 | EXPORT_SYMBOL(full_name_hash); | |
1998 | ||
fcfd2fbf | 1999 | /* Return the "hash_len" (hash and length) of a null-terminated string */ |
8387ff25 | 2000 | u64 hashlen_string(const void *salt, const char *name) |
fcfd2fbf | 2001 | { |
8387ff25 LT |
2002 | unsigned long a = 0, x = 0, y = (unsigned long)salt; |
2003 | unsigned long adata, mask, len; | |
fcfd2fbf GS |
2004 | const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS; |
2005 | ||
8387ff25 LT |
2006 | len = 0; |
2007 | goto inside; | |
2008 | ||
fcfd2fbf | 2009 | do { |
2a18da7a | 2010 | HASH_MIX(x, y, a); |
fcfd2fbf | 2011 | len += sizeof(unsigned long); |
8387ff25 | 2012 | inside: |
fcfd2fbf GS |
2013 | a = load_unaligned_zeropad(name+len); |
2014 | } while (!has_zero(a, &adata, &constants)); | |
2015 | ||
2016 | adata = prep_zero_mask(a, adata, &constants); | |
2017 | mask = create_zero_mask(adata); | |
2a18da7a | 2018 | x ^= a & zero_bytemask(mask); |
fcfd2fbf | 2019 | |
2a18da7a | 2020 | return hashlen_create(fold_hash(x, y), len + find_zero(mask)); |
fcfd2fbf GS |
2021 | } |
2022 | EXPORT_SYMBOL(hashlen_string); | |
2023 | ||
bfcfaa77 LT |
2024 | /* |
2025 | * Calculate the length and hash of the path component, and | |
d6bb3e90 | 2026 | * return the "hash_len" as the result. |
bfcfaa77 | 2027 | */ |
8387ff25 | 2028 | static inline u64 hash_name(const void *salt, const char *name) |
bfcfaa77 | 2029 | { |
8387ff25 LT |
2030 | unsigned long a = 0, b, x = 0, y = (unsigned long)salt; |
2031 | unsigned long adata, bdata, mask, len; | |
36126f8f | 2032 | const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS; |
bfcfaa77 | 2033 | |
8387ff25 LT |
2034 | len = 0; |
2035 | goto inside; | |
2036 | ||
bfcfaa77 | 2037 | do { |
2a18da7a | 2038 | HASH_MIX(x, y, a); |
bfcfaa77 | 2039 | len += sizeof(unsigned long); |
8387ff25 | 2040 | inside: |
e419b4cc | 2041 | a = load_unaligned_zeropad(name+len); |
36126f8f LT |
2042 | b = a ^ REPEAT_BYTE('/'); |
2043 | } while (!(has_zero(a, &adata, &constants) | has_zero(b, &bdata, &constants))); | |
2044 | ||
2045 | adata = prep_zero_mask(a, adata, &constants); | |
2046 | bdata = prep_zero_mask(b, bdata, &constants); | |
36126f8f | 2047 | mask = create_zero_mask(adata | bdata); |
2a18da7a | 2048 | x ^= a & zero_bytemask(mask); |
36126f8f | 2049 | |
2a18da7a | 2050 | return hashlen_create(fold_hash(x, y), len + find_zero(mask)); |
bfcfaa77 LT |
2051 | } |
2052 | ||
2a18da7a | 2053 | #else /* !CONFIG_DCACHE_WORD_ACCESS: Slow, byte-at-a-time version */ |
bfcfaa77 | 2054 | |
fcfd2fbf | 2055 | /* Return the hash of a string of known length */ |
8387ff25 | 2056 | unsigned int full_name_hash(const void *salt, const char *name, unsigned int len) |
0145acc2 | 2057 | { |
8387ff25 | 2058 | unsigned long hash = init_name_hash(salt); |
0145acc2 | 2059 | while (len--) |
fcfd2fbf | 2060 | hash = partial_name_hash((unsigned char)*name++, hash); |
0145acc2 LT |
2061 | return end_name_hash(hash); |
2062 | } | |
ae942ae7 | 2063 | EXPORT_SYMBOL(full_name_hash); |
0145acc2 | 2064 | |
fcfd2fbf | 2065 | /* Return the "hash_len" (hash and length) of a null-terminated string */ |
8387ff25 | 2066 | u64 hashlen_string(const void *salt, const char *name) |
fcfd2fbf | 2067 | { |
8387ff25 | 2068 | unsigned long hash = init_name_hash(salt); |
fcfd2fbf GS |
2069 | unsigned long len = 0, c; |
2070 | ||
2071 | c = (unsigned char)*name; | |
e0ab7af9 | 2072 | while (c) { |
fcfd2fbf GS |
2073 | len++; |
2074 | hash = partial_name_hash(c, hash); | |
2075 | c = (unsigned char)name[len]; | |
e0ab7af9 | 2076 | } |
fcfd2fbf GS |
2077 | return hashlen_create(end_name_hash(hash), len); |
2078 | } | |
f2a031b6 | 2079 | EXPORT_SYMBOL(hashlen_string); |
fcfd2fbf | 2080 | |
200e9ef7 LT |
2081 | /* |
2082 | * We know there's a real path component here of at least | |
2083 | * one character. | |
2084 | */ | |
8387ff25 | 2085 | static inline u64 hash_name(const void *salt, const char *name) |
200e9ef7 | 2086 | { |
8387ff25 | 2087 | unsigned long hash = init_name_hash(salt); |
200e9ef7 LT |
2088 | unsigned long len = 0, c; |
2089 | ||
2090 | c = (unsigned char)*name; | |
2091 | do { | |
2092 | len++; | |
2093 | hash = partial_name_hash(c, hash); | |
2094 | c = (unsigned char)name[len]; | |
2095 | } while (c && c != '/'); | |
d6bb3e90 | 2096 | return hashlen_create(end_name_hash(hash), len); |
200e9ef7 LT |
2097 | } |
2098 | ||
bfcfaa77 LT |
2099 | #endif |
2100 | ||
1da177e4 LT |
2101 | /* |
2102 | * Name resolution. | |
ea3834d9 PM |
2103 | * This is the basic name resolution function, turning a pathname into |
2104 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 2105 | * |
ea3834d9 PM |
2106 | * Returns 0 and nd will have valid dentry and mnt on success. |
2107 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 2108 | */ |
6de88d72 | 2109 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 | 2110 | { |
d8d4611a | 2111 | int depth = 0; // depth <= nd->depth |
1da177e4 | 2112 | int err; |
32cd7468 | 2113 | |
b4c03536 | 2114 | nd->last_type = LAST_ROOT; |
c108837e | 2115 | nd->flags |= LOOKUP_PARENT; |
9b5858e9 AV |
2116 | if (IS_ERR(name)) |
2117 | return PTR_ERR(name); | |
1da177e4 LT |
2118 | while (*name=='/') |
2119 | name++; | |
1a97d899 AV |
2120 | if (!*name) { |
2121 | nd->dir_mode = 0; // short-circuit the 'hardening' idiocy | |
9e18f10a | 2122 | return 0; |
1a97d899 | 2123 | } |
1da177e4 | 2124 | |
1da177e4 LT |
2125 | /* At this point we know we have a real path component. */ |
2126 | for(;;) { | |
92d27016 | 2127 | const char *link; |
d6bb3e90 | 2128 | u64 hash_len; |
fe479a58 | 2129 | int type; |
1da177e4 | 2130 | |
52094c8a | 2131 | err = may_lookup(nd); |
2a18da7a | 2132 | if (err) |
3595e234 | 2133 | return err; |
1da177e4 | 2134 | |
8387ff25 | 2135 | hash_len = hash_name(nd->path.dentry, name); |
1da177e4 | 2136 | |
fe479a58 | 2137 | type = LAST_NORM; |
d6bb3e90 | 2138 | if (name[0] == '.') switch (hashlen_len(hash_len)) { |
fe479a58 | 2139 | case 2: |
200e9ef7 | 2140 | if (name[1] == '.') { |
fe479a58 | 2141 | type = LAST_DOTDOT; |
16c2cd71 AV |
2142 | nd->flags |= LOOKUP_JUMPED; |
2143 | } | |
fe479a58 AV |
2144 | break; |
2145 | case 1: | |
2146 | type = LAST_DOT; | |
2147 | } | |
5a202bcd AV |
2148 | if (likely(type == LAST_NORM)) { |
2149 | struct dentry *parent = nd->path.dentry; | |
16c2cd71 | 2150 | nd->flags &= ~LOOKUP_JUMPED; |
5a202bcd | 2151 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
a060dc50 | 2152 | struct qstr this = { { .hash_len = hash_len }, .name = name }; |
da53be12 | 2153 | err = parent->d_op->d_hash(parent, &this); |
5a202bcd | 2154 | if (err < 0) |
3595e234 | 2155 | return err; |
d6bb3e90 LT |
2156 | hash_len = this.hash_len; |
2157 | name = this.name; | |
5a202bcd AV |
2158 | } |
2159 | } | |
fe479a58 | 2160 | |
d6bb3e90 LT |
2161 | nd->last.hash_len = hash_len; |
2162 | nd->last.name = name; | |
5f4a6a69 AV |
2163 | nd->last_type = type; |
2164 | ||
d6bb3e90 LT |
2165 | name += hashlen_len(hash_len); |
2166 | if (!*name) | |
bdf6cbf1 | 2167 | goto OK; |
200e9ef7 LT |
2168 | /* |
2169 | * If it wasn't NUL, we know it was '/'. Skip that | |
2170 | * slash, and continue until no more slashes. | |
2171 | */ | |
2172 | do { | |
d6bb3e90 LT |
2173 | name++; |
2174 | } while (unlikely(*name == '/')); | |
8620c238 AV |
2175 | if (unlikely(!*name)) { |
2176 | OK: | |
d8d4611a | 2177 | /* pathname or trailing symlink, done */ |
c108837e | 2178 | if (!depth) { |
0f705953 AV |
2179 | nd->dir_uid = nd->inode->i_uid; |
2180 | nd->dir_mode = nd->inode->i_mode; | |
c108837e | 2181 | nd->flags &= ~LOOKUP_PARENT; |
8620c238 | 2182 | return 0; |
c108837e | 2183 | } |
8620c238 | 2184 | /* last component of nested symlink */ |
d8d4611a | 2185 | name = nd->stack[--depth].name; |
8c4efe22 | 2186 | link = walk_component(nd, 0); |
1c4ff1a8 AV |
2187 | } else { |
2188 | /* not the last component */ | |
8c4efe22 | 2189 | link = walk_component(nd, WALK_MORE); |
8620c238 | 2190 | } |
92d27016 AV |
2191 | if (unlikely(link)) { |
2192 | if (IS_ERR(link)) | |
2193 | return PTR_ERR(link); | |
2194 | /* a symlink to follow */ | |
d8d4611a | 2195 | nd->stack[depth++].name = name; |
92d27016 AV |
2196 | name = link; |
2197 | continue; | |
31e6b01f | 2198 | } |
97242f99 AV |
2199 | if (unlikely(!d_can_lookup(nd->path.dentry))) { |
2200 | if (nd->flags & LOOKUP_RCU) { | |
4675ac39 | 2201 | if (unlazy_walk(nd)) |
97242f99 AV |
2202 | return -ECHILD; |
2203 | } | |
3595e234 | 2204 | return -ENOTDIR; |
97242f99 | 2205 | } |
1da177e4 | 2206 | } |
1da177e4 LT |
2207 | } |
2208 | ||
edc2b1da | 2209 | /* must be paired with terminate_walk() */ |
c8a53ee5 | 2210 | static const char *path_init(struct nameidata *nd, unsigned flags) |
31e6b01f | 2211 | { |
740a1678 | 2212 | int error; |
c8a53ee5 | 2213 | const char *s = nd->name->name; |
31e6b01f | 2214 | |
c0eb027e LT |
2215 | if (!*s) |
2216 | flags &= ~LOOKUP_RCU; | |
edc2b1da AV |
2217 | if (flags & LOOKUP_RCU) |
2218 | rcu_read_lock(); | |
c0eb027e | 2219 | |
c108837e | 2220 | nd->flags = flags | LOOKUP_JUMPED; |
31e6b01f | 2221 | nd->depth = 0; |
ab87f9a5 AS |
2222 | |
2223 | nd->m_seq = __read_seqcount_begin(&mount_lock.seqcount); | |
2224 | nd->r_seq = __read_seqcount_begin(&rename_lock.seqcount); | |
2225 | smp_rmb(); | |
2226 | ||
5b6ca027 | 2227 | if (flags & LOOKUP_ROOT) { |
b18825a7 DH |
2228 | struct dentry *root = nd->root.dentry; |
2229 | struct inode *inode = root->d_inode; | |
93893862 AV |
2230 | if (*s && unlikely(!d_can_lookup(root))) |
2231 | return ERR_PTR(-ENOTDIR); | |
5b6ca027 AV |
2232 | nd->path = nd->root; |
2233 | nd->inode = inode; | |
2234 | if (flags & LOOKUP_RCU) { | |
ab87f9a5 | 2235 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); |
8f47a016 | 2236 | nd->root_seq = nd->seq; |
5b6ca027 AV |
2237 | } else { |
2238 | path_get(&nd->path); | |
2239 | } | |
368ee9ba | 2240 | return s; |
5b6ca027 AV |
2241 | } |
2242 | ||
31e6b01f | 2243 | nd->root.mnt = NULL; |
248fb5b9 AV |
2244 | nd->path.mnt = NULL; |
2245 | nd->path.dentry = NULL; | |
31e6b01f | 2246 | |
8db52c7e AS |
2247 | /* Absolute pathname -- fetch the root (LOOKUP_IN_ROOT uses nd->dfd). */ |
2248 | if (*s == '/' && !(flags & LOOKUP_IN_ROOT)) { | |
740a1678 AS |
2249 | error = nd_jump_root(nd); |
2250 | if (unlikely(error)) | |
2251 | return ERR_PTR(error); | |
2252 | return s; | |
8db52c7e AS |
2253 | } |
2254 | ||
2255 | /* Relative pathname -- get the starting-point it is relative to. */ | |
2256 | if (nd->dfd == AT_FDCWD) { | |
e41f7d4e AV |
2257 | if (flags & LOOKUP_RCU) { |
2258 | struct fs_struct *fs = current->fs; | |
2259 | unsigned seq; | |
31e6b01f | 2260 | |
e41f7d4e AV |
2261 | do { |
2262 | seq = read_seqcount_begin(&fs->seq); | |
2263 | nd->path = fs->pwd; | |
ef55d917 | 2264 | nd->inode = nd->path.dentry->d_inode; |
e41f7d4e AV |
2265 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
2266 | } while (read_seqcount_retry(&fs->seq, seq)); | |
2267 | } else { | |
2268 | get_fs_pwd(current->fs, &nd->path); | |
ef55d917 | 2269 | nd->inode = nd->path.dentry->d_inode; |
e41f7d4e | 2270 | } |
31e6b01f | 2271 | } else { |
582aa64a | 2272 | /* Caller must check execute permissions on the starting path component */ |
c8a53ee5 | 2273 | struct fd f = fdget_raw(nd->dfd); |
31e6b01f NP |
2274 | struct dentry *dentry; |
2275 | ||
2903ff01 | 2276 | if (!f.file) |
368ee9ba | 2277 | return ERR_PTR(-EBADF); |
31e6b01f | 2278 | |
2903ff01 | 2279 | dentry = f.file->f_path.dentry; |
31e6b01f | 2280 | |
edc2b1da AV |
2281 | if (*s && unlikely(!d_can_lookup(dentry))) { |
2282 | fdput(f); | |
2283 | return ERR_PTR(-ENOTDIR); | |
f52e0c11 | 2284 | } |
31e6b01f | 2285 | |
2903ff01 | 2286 | nd->path = f.file->f_path; |
e41f7d4e | 2287 | if (flags & LOOKUP_RCU) { |
34a26b99 AV |
2288 | nd->inode = nd->path.dentry->d_inode; |
2289 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
e41f7d4e | 2290 | } else { |
2903ff01 | 2291 | path_get(&nd->path); |
34a26b99 | 2292 | nd->inode = nd->path.dentry->d_inode; |
e41f7d4e | 2293 | } |
34a26b99 | 2294 | fdput(f); |
31e6b01f | 2295 | } |
8db52c7e | 2296 | |
adb21d2b AS |
2297 | /* For scoped-lookups we need to set the root to the dirfd as well. */ |
2298 | if (flags & LOOKUP_IS_SCOPED) { | |
2299 | nd->root = nd->path; | |
2300 | if (flags & LOOKUP_RCU) { | |
2301 | nd->root_seq = nd->seq; | |
2302 | } else { | |
2303 | path_get(&nd->root); | |
2304 | nd->flags |= LOOKUP_ROOT_GRABBED; | |
2305 | } | |
2306 | } | |
2307 | return s; | |
9b4a9b14 AV |
2308 | } |
2309 | ||
1ccac622 | 2310 | static inline const char *lookup_last(struct nameidata *nd) |
bd92d7fe AV |
2311 | { |
2312 | if (nd->last_type == LAST_NORM && nd->last.name[nd->last.len]) | |
2313 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
2314 | ||
c108837e | 2315 | return walk_component(nd, WALK_TRAILING); |
bd92d7fe AV |
2316 | } |
2317 | ||
4f757f3c AV |
2318 | static int handle_lookup_down(struct nameidata *nd) |
2319 | { | |
c153007b | 2320 | if (!(nd->flags & LOOKUP_RCU)) |
db3c9ade | 2321 | dget(nd->path.dentry); |
b0417d2c AV |
2322 | return PTR_ERR(step_into(nd, WALK_NOFOLLOW, |
2323 | nd->path.dentry, nd->inode, nd->seq)); | |
4f757f3c AV |
2324 | } |
2325 | ||
9b4a9b14 | 2326 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
c8a53ee5 | 2327 | static int path_lookupat(struct nameidata *nd, unsigned flags, struct path *path) |
9b4a9b14 | 2328 | { |
c8a53ee5 | 2329 | const char *s = path_init(nd, flags); |
bd92d7fe | 2330 | int err; |
31e6b01f | 2331 | |
9b5858e9 | 2332 | if (unlikely(flags & LOOKUP_DOWN) && !IS_ERR(s)) { |
4f757f3c | 2333 | err = handle_lookup_down(nd); |
5f336e72 AV |
2334 | if (unlikely(err < 0)) |
2335 | s = ERR_PTR(err); | |
4f757f3c AV |
2336 | } |
2337 | ||
1ccac622 AV |
2338 | while (!(err = link_path_walk(s, nd)) && |
2339 | (s = lookup_last(nd)) != NULL) | |
2340 | ; | |
9f1fafee AV |
2341 | if (!err) |
2342 | err = complete_walk(nd); | |
bd92d7fe | 2343 | |
deb106c6 AV |
2344 | if (!err && nd->flags & LOOKUP_DIRECTORY) |
2345 | if (!d_can_lookup(nd->path.dentry)) | |
bd23a539 | 2346 | err = -ENOTDIR; |
161aff1d AV |
2347 | if (!err && unlikely(nd->flags & LOOKUP_MOUNTPOINT)) { |
2348 | err = handle_lookup_down(nd); | |
2349 | nd->flags &= ~LOOKUP_JUMPED; // no d_weak_revalidate(), please... | |
2350 | } | |
625b6d10 AV |
2351 | if (!err) { |
2352 | *path = nd->path; | |
2353 | nd->path.mnt = NULL; | |
2354 | nd->path.dentry = NULL; | |
2355 | } | |
2356 | terminate_walk(nd); | |
bd92d7fe | 2357 | return err; |
ee0827cd | 2358 | } |
31e6b01f | 2359 | |
31d921c7 DH |
2360 | int filename_lookup(int dfd, struct filename *name, unsigned flags, |
2361 | struct path *path, struct path *root) | |
ee0827cd | 2362 | { |
894bc8c4 | 2363 | int retval; |
9883d185 | 2364 | struct nameidata nd; |
abc9f5be AV |
2365 | if (IS_ERR(name)) |
2366 | return PTR_ERR(name); | |
9ad1aaa6 AV |
2367 | if (unlikely(root)) { |
2368 | nd.root = *root; | |
2369 | flags |= LOOKUP_ROOT; | |
2370 | } | |
9883d185 | 2371 | set_nameidata(&nd, dfd, name); |
c8a53ee5 | 2372 | retval = path_lookupat(&nd, flags | LOOKUP_RCU, path); |
ee0827cd | 2373 | if (unlikely(retval == -ECHILD)) |
c8a53ee5 | 2374 | retval = path_lookupat(&nd, flags, path); |
ee0827cd | 2375 | if (unlikely(retval == -ESTALE)) |
c8a53ee5 | 2376 | retval = path_lookupat(&nd, flags | LOOKUP_REVAL, path); |
31e6b01f | 2377 | |
f78570dd | 2378 | if (likely(!retval)) |
161aff1d AV |
2379 | audit_inode(name, path->dentry, |
2380 | flags & LOOKUP_MOUNTPOINT ? AUDIT_INODE_NOEVAL : 0); | |
9883d185 | 2381 | restore_nameidata(); |
e4bd1c1a | 2382 | putname(name); |
170aa3d0 | 2383 | return retval; |
1da177e4 LT |
2384 | } |
2385 | ||
8bcb77fa | 2386 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
c8a53ee5 | 2387 | static int path_parentat(struct nameidata *nd, unsigned flags, |
391172c4 | 2388 | struct path *parent) |
8bcb77fa | 2389 | { |
c8a53ee5 | 2390 | const char *s = path_init(nd, flags); |
9b5858e9 | 2391 | int err = link_path_walk(s, nd); |
8bcb77fa AV |
2392 | if (!err) |
2393 | err = complete_walk(nd); | |
391172c4 AV |
2394 | if (!err) { |
2395 | *parent = nd->path; | |
2396 | nd->path.mnt = NULL; | |
2397 | nd->path.dentry = NULL; | |
2398 | } | |
2399 | terminate_walk(nd); | |
8bcb77fa AV |
2400 | return err; |
2401 | } | |
2402 | ||
5c31b6ce | 2403 | static struct filename *filename_parentat(int dfd, struct filename *name, |
391172c4 AV |
2404 | unsigned int flags, struct path *parent, |
2405 | struct qstr *last, int *type) | |
8bcb77fa AV |
2406 | { |
2407 | int retval; | |
9883d185 | 2408 | struct nameidata nd; |
8bcb77fa | 2409 | |
5c31b6ce AV |
2410 | if (IS_ERR(name)) |
2411 | return name; | |
9883d185 | 2412 | set_nameidata(&nd, dfd, name); |
c8a53ee5 | 2413 | retval = path_parentat(&nd, flags | LOOKUP_RCU, parent); |
8bcb77fa | 2414 | if (unlikely(retval == -ECHILD)) |
c8a53ee5 | 2415 | retval = path_parentat(&nd, flags, parent); |
8bcb77fa | 2416 | if (unlikely(retval == -ESTALE)) |
c8a53ee5 | 2417 | retval = path_parentat(&nd, flags | LOOKUP_REVAL, parent); |
391172c4 AV |
2418 | if (likely(!retval)) { |
2419 | *last = nd.last; | |
2420 | *type = nd.last_type; | |
c9b07eab | 2421 | audit_inode(name, parent->dentry, AUDIT_INODE_PARENT); |
5c31b6ce AV |
2422 | } else { |
2423 | putname(name); | |
2424 | name = ERR_PTR(retval); | |
391172c4 | 2425 | } |
9883d185 | 2426 | restore_nameidata(); |
5c31b6ce | 2427 | return name; |
8bcb77fa AV |
2428 | } |
2429 | ||
79714f72 AV |
2430 | /* does lookup, returns the object with parent locked */ |
2431 | struct dentry *kern_path_locked(const char *name, struct path *path) | |
5590ff0d | 2432 | { |
5c31b6ce AV |
2433 | struct filename *filename; |
2434 | struct dentry *d; | |
391172c4 AV |
2435 | struct qstr last; |
2436 | int type; | |
51689104 | 2437 | |
5c31b6ce AV |
2438 | filename = filename_parentat(AT_FDCWD, getname_kernel(name), 0, path, |
2439 | &last, &type); | |
51689104 PM |
2440 | if (IS_ERR(filename)) |
2441 | return ERR_CAST(filename); | |
5c31b6ce | 2442 | if (unlikely(type != LAST_NORM)) { |
391172c4 | 2443 | path_put(path); |
5c31b6ce AV |
2444 | putname(filename); |
2445 | return ERR_PTR(-EINVAL); | |
79714f72 | 2446 | } |
5955102c | 2447 | inode_lock_nested(path->dentry->d_inode, I_MUTEX_PARENT); |
391172c4 | 2448 | d = __lookup_hash(&last, path->dentry, 0); |
79714f72 | 2449 | if (IS_ERR(d)) { |
5955102c | 2450 | inode_unlock(path->dentry->d_inode); |
391172c4 | 2451 | path_put(path); |
79714f72 | 2452 | } |
51689104 | 2453 | putname(filename); |
79714f72 | 2454 | return d; |
5590ff0d UD |
2455 | } |
2456 | ||
d1811465 AV |
2457 | int kern_path(const char *name, unsigned int flags, struct path *path) |
2458 | { | |
abc9f5be AV |
2459 | return filename_lookup(AT_FDCWD, getname_kernel(name), |
2460 | flags, path, NULL); | |
d1811465 | 2461 | } |
4d359507 | 2462 | EXPORT_SYMBOL(kern_path); |
d1811465 | 2463 | |
16f18200 JJS |
2464 | /** |
2465 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
2466 | * @dentry: pointer to dentry of the base directory | |
2467 | * @mnt: pointer to vfs mount of the base directory | |
2468 | * @name: pointer to file name | |
2469 | * @flags: lookup flags | |
e0a01249 | 2470 | * @path: pointer to struct path to fill |
16f18200 JJS |
2471 | */ |
2472 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
2473 | const char *name, unsigned int flags, | |
e0a01249 | 2474 | struct path *path) |
16f18200 | 2475 | { |
9ad1aaa6 | 2476 | struct path root = {.mnt = mnt, .dentry = dentry}; |
9ad1aaa6 | 2477 | /* the first argument of filename_lookup() is ignored with root */ |
abc9f5be AV |
2478 | return filename_lookup(AT_FDCWD, getname_kernel(name), |
2479 | flags , path, &root); | |
16f18200 | 2480 | } |
4d359507 | 2481 | EXPORT_SYMBOL(vfs_path_lookup); |
16f18200 | 2482 | |
3c95f0dc AV |
2483 | static int lookup_one_len_common(const char *name, struct dentry *base, |
2484 | int len, struct qstr *this) | |
057f6c01 | 2485 | { |
3c95f0dc AV |
2486 | this->name = name; |
2487 | this->len = len; | |
2488 | this->hash = full_name_hash(base, name, len); | |
6a96ba54 | 2489 | if (!len) |
3c95f0dc | 2490 | return -EACCES; |
6a96ba54 | 2491 | |
21d8a15a AV |
2492 | if (unlikely(name[0] == '.')) { |
2493 | if (len < 2 || (len == 2 && name[1] == '.')) | |
3c95f0dc | 2494 | return -EACCES; |
21d8a15a AV |
2495 | } |
2496 | ||
6a96ba54 | 2497 | while (len--) { |
3c95f0dc | 2498 | unsigned int c = *(const unsigned char *)name++; |
6a96ba54 | 2499 | if (c == '/' || c == '\0') |
3c95f0dc | 2500 | return -EACCES; |
6a96ba54 | 2501 | } |
5a202bcd AV |
2502 | /* |
2503 | * See if the low-level filesystem might want | |
2504 | * to use its own hash.. | |
2505 | */ | |
2506 | if (base->d_flags & DCACHE_OP_HASH) { | |
3c95f0dc | 2507 | int err = base->d_op->d_hash(base, this); |
5a202bcd | 2508 | if (err < 0) |
3c95f0dc | 2509 | return err; |
5a202bcd | 2510 | } |
eead1911 | 2511 | |
3c95f0dc AV |
2512 | return inode_permission(base->d_inode, MAY_EXEC); |
2513 | } | |
2514 | ||
0da0b7fd DH |
2515 | /** |
2516 | * try_lookup_one_len - filesystem helper to lookup single pathname component | |
2517 | * @name: pathname component to lookup | |
2518 | * @base: base directory to lookup from | |
2519 | * @len: maximum length @len should be interpreted to | |
2520 | * | |
2521 | * Look up a dentry by name in the dcache, returning NULL if it does not | |
2522 | * currently exist. The function does not try to create a dentry. | |
2523 | * | |
2524 | * Note that this routine is purely a helper for filesystem usage and should | |
2525 | * not be called by generic code. | |
2526 | * | |
2527 | * The caller must hold base->i_mutex. | |
2528 | */ | |
2529 | struct dentry *try_lookup_one_len(const char *name, struct dentry *base, int len) | |
2530 | { | |
2531 | struct qstr this; | |
2532 | int err; | |
2533 | ||
2534 | WARN_ON_ONCE(!inode_is_locked(base->d_inode)); | |
2535 | ||
2536 | err = lookup_one_len_common(name, base, len, &this); | |
2537 | if (err) | |
2538 | return ERR_PTR(err); | |
2539 | ||
2540 | return lookup_dcache(&this, base, 0); | |
2541 | } | |
2542 | EXPORT_SYMBOL(try_lookup_one_len); | |
2543 | ||
3c95f0dc AV |
2544 | /** |
2545 | * lookup_one_len - filesystem helper to lookup single pathname component | |
2546 | * @name: pathname component to lookup | |
2547 | * @base: base directory to lookup from | |
2548 | * @len: maximum length @len should be interpreted to | |
2549 | * | |
2550 | * Note that this routine is purely a helper for filesystem usage and should | |
2551 | * not be called by generic code. | |
2552 | * | |
2553 | * The caller must hold base->i_mutex. | |
2554 | */ | |
2555 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) | |
2556 | { | |
8613a209 | 2557 | struct dentry *dentry; |
3c95f0dc AV |
2558 | struct qstr this; |
2559 | int err; | |
2560 | ||
2561 | WARN_ON_ONCE(!inode_is_locked(base->d_inode)); | |
2562 | ||
2563 | err = lookup_one_len_common(name, base, len, &this); | |
cda309de MS |
2564 | if (err) |
2565 | return ERR_PTR(err); | |
2566 | ||
8613a209 AV |
2567 | dentry = lookup_dcache(&this, base, 0); |
2568 | return dentry ? dentry : __lookup_slow(&this, base, 0); | |
057f6c01 | 2569 | } |
4d359507 | 2570 | EXPORT_SYMBOL(lookup_one_len); |
057f6c01 | 2571 | |
bbddca8e N |
2572 | /** |
2573 | * lookup_one_len_unlocked - filesystem helper to lookup single pathname component | |
2574 | * @name: pathname component to lookup | |
2575 | * @base: base directory to lookup from | |
2576 | * @len: maximum length @len should be interpreted to | |
2577 | * | |
2578 | * Note that this routine is purely a helper for filesystem usage and should | |
2579 | * not be called by generic code. | |
2580 | * | |
2581 | * Unlike lookup_one_len, it should be called without the parent | |
2582 | * i_mutex held, and will take the i_mutex itself if necessary. | |
2583 | */ | |
2584 | struct dentry *lookup_one_len_unlocked(const char *name, | |
2585 | struct dentry *base, int len) | |
2586 | { | |
2587 | struct qstr this; | |
bbddca8e | 2588 | int err; |
20d00ee8 | 2589 | struct dentry *ret; |
bbddca8e | 2590 | |
3c95f0dc | 2591 | err = lookup_one_len_common(name, base, len, &this); |
bbddca8e N |
2592 | if (err) |
2593 | return ERR_PTR(err); | |
2594 | ||
20d00ee8 LT |
2595 | ret = lookup_dcache(&this, base, 0); |
2596 | if (!ret) | |
2597 | ret = lookup_slow(&this, base, 0); | |
2598 | return ret; | |
bbddca8e N |
2599 | } |
2600 | EXPORT_SYMBOL(lookup_one_len_unlocked); | |
2601 | ||
6c2d4798 AV |
2602 | /* |
2603 | * Like lookup_one_len_unlocked(), except that it yields ERR_PTR(-ENOENT) | |
2604 | * on negatives. Returns known positive or ERR_PTR(); that's what | |
2605 | * most of the users want. Note that pinned negative with unlocked parent | |
2606 | * _can_ become positive at any time, so callers of lookup_one_len_unlocked() | |
2607 | * need to be very careful; pinned positives have ->d_inode stable, so | |
2608 | * this one avoids such problems. | |
2609 | */ | |
2610 | struct dentry *lookup_positive_unlocked(const char *name, | |
2611 | struct dentry *base, int len) | |
2612 | { | |
2613 | struct dentry *ret = lookup_one_len_unlocked(name, base, len); | |
2fa6b1e0 | 2614 | if (!IS_ERR(ret) && d_flags_negative(smp_load_acquire(&ret->d_flags))) { |
6c2d4798 AV |
2615 | dput(ret); |
2616 | ret = ERR_PTR(-ENOENT); | |
2617 | } | |
2618 | return ret; | |
2619 | } | |
2620 | EXPORT_SYMBOL(lookup_positive_unlocked); | |
2621 | ||
eedf265a EB |
2622 | #ifdef CONFIG_UNIX98_PTYS |
2623 | int path_pts(struct path *path) | |
2624 | { | |
2625 | /* Find something mounted on "pts" in the same directory as | |
2626 | * the input path. | |
2627 | */ | |
a6a7eb76 AV |
2628 | struct dentry *parent = dget_parent(path->dentry); |
2629 | struct dentry *child; | |
19f6028a | 2630 | struct qstr this = QSTR_INIT("pts", 3); |
eedf265a | 2631 | |
a6a7eb76 AV |
2632 | if (unlikely(!path_connected(path->mnt, parent))) { |
2633 | dput(parent); | |
63b27720 | 2634 | return -ENOENT; |
a6a7eb76 | 2635 | } |
63b27720 AV |
2636 | dput(path->dentry); |
2637 | path->dentry = parent; | |
eedf265a EB |
2638 | child = d_hash_and_lookup(parent, &this); |
2639 | if (!child) | |
2640 | return -ENOENT; | |
2641 | ||
2642 | path->dentry = child; | |
2643 | dput(parent); | |
19f6028a | 2644 | follow_down(path); |
eedf265a EB |
2645 | return 0; |
2646 | } | |
2647 | #endif | |
2648 | ||
1fa1e7f6 AW |
2649 | int user_path_at_empty(int dfd, const char __user *name, unsigned flags, |
2650 | struct path *path, int *empty) | |
1da177e4 | 2651 | { |
abc9f5be AV |
2652 | return filename_lookup(dfd, getname_flags(name, flags, empty), |
2653 | flags, path, NULL); | |
1da177e4 | 2654 | } |
b853a161 | 2655 | EXPORT_SYMBOL(user_path_at_empty); |
1fa1e7f6 | 2656 | |
cbdf35bc | 2657 | int __check_sticky(struct inode *dir, struct inode *inode) |
1da177e4 | 2658 | { |
8e96e3b7 | 2659 | kuid_t fsuid = current_fsuid(); |
da9592ed | 2660 | |
8e96e3b7 | 2661 | if (uid_eq(inode->i_uid, fsuid)) |
1da177e4 | 2662 | return 0; |
8e96e3b7 | 2663 | if (uid_eq(dir->i_uid, fsuid)) |
1da177e4 | 2664 | return 0; |
0558c1bf | 2665 | return !capable_wrt_inode_uidgid(&init_user_ns, inode, CAP_FOWNER); |
1da177e4 | 2666 | } |
cbdf35bc | 2667 | EXPORT_SYMBOL(__check_sticky); |
1da177e4 LT |
2668 | |
2669 | /* | |
2670 | * Check whether we can remove a link victim from directory dir, check | |
2671 | * whether the type of victim is right. | |
2672 | * 1. We can't do it if dir is read-only (done in permission()) | |
2673 | * 2. We should have write and exec permissions on dir | |
2674 | * 3. We can't remove anything from append-only dir | |
2675 | * 4. We can't do anything with immutable dir (done in permission()) | |
2676 | * 5. If the sticky bit on dir is set we should either | |
2677 | * a. be owner of dir, or | |
2678 | * b. be owner of victim, or | |
2679 | * c. have CAP_FOWNER capability | |
2680 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
2681 | * links pointing to it. | |
0bd23d09 EB |
2682 | * 7. If the victim has an unknown uid or gid we can't change the inode. |
2683 | * 8. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
2684 | * 9. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
2685 | * 10. We can't remove a root or mountpoint. | |
2686 | * 11. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1da177e4 LT |
2687 | * nfs_async_unlink(). |
2688 | */ | |
b18825a7 | 2689 | static int may_delete(struct inode *dir, struct dentry *victim, bool isdir) |
1da177e4 | 2690 | { |
63afdfc7 | 2691 | struct inode *inode = d_backing_inode(victim); |
1da177e4 LT |
2692 | int error; |
2693 | ||
b18825a7 | 2694 | if (d_is_negative(victim)) |
1da177e4 | 2695 | return -ENOENT; |
b18825a7 | 2696 | BUG_ON(!inode); |
1da177e4 LT |
2697 | |
2698 | BUG_ON(victim->d_parent->d_inode != dir); | |
593d1ce8 EB |
2699 | |
2700 | /* Inode writeback is not safe when the uid or gid are invalid. */ | |
2701 | if (!uid_valid(inode->i_uid) || !gid_valid(inode->i_gid)) | |
2702 | return -EOVERFLOW; | |
2703 | ||
4fa6b5ec | 2704 | audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); |
1da177e4 | 2705 | |
f419a2e3 | 2706 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2707 | if (error) |
2708 | return error; | |
2709 | if (IS_APPEND(dir)) | |
2710 | return -EPERM; | |
b18825a7 DH |
2711 | |
2712 | if (check_sticky(dir, inode) || IS_APPEND(inode) || | |
0bd23d09 | 2713 | IS_IMMUTABLE(inode) || IS_SWAPFILE(inode) || HAS_UNMAPPED_ID(inode)) |
1da177e4 LT |
2714 | return -EPERM; |
2715 | if (isdir) { | |
44b1d530 | 2716 | if (!d_is_dir(victim)) |
1da177e4 LT |
2717 | return -ENOTDIR; |
2718 | if (IS_ROOT(victim)) | |
2719 | return -EBUSY; | |
44b1d530 | 2720 | } else if (d_is_dir(victim)) |
1da177e4 LT |
2721 | return -EISDIR; |
2722 | if (IS_DEADDIR(dir)) | |
2723 | return -ENOENT; | |
2724 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
2725 | return -EBUSY; | |
2726 | return 0; | |
2727 | } | |
2728 | ||
2729 | /* Check whether we can create an object with dentry child in directory | |
2730 | * dir. | |
2731 | * 1. We can't do it if child already exists (open has special treatment for | |
2732 | * this case, but since we are inlined it's OK) | |
2733 | * 2. We can't do it if dir is read-only (done in permission()) | |
036d5236 EB |
2734 | * 3. We can't do it if the fs can't represent the fsuid or fsgid. |
2735 | * 4. We should have write and exec permissions on dir | |
2736 | * 5. We can't do it if dir is immutable (done in permission()) | |
1da177e4 | 2737 | */ |
a95164d9 | 2738 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 | 2739 | { |
036d5236 | 2740 | struct user_namespace *s_user_ns; |
14e972b4 | 2741 | audit_inode_child(dir, child, AUDIT_TYPE_CHILD_CREATE); |
1da177e4 LT |
2742 | if (child->d_inode) |
2743 | return -EEXIST; | |
2744 | if (IS_DEADDIR(dir)) | |
2745 | return -ENOENT; | |
036d5236 EB |
2746 | s_user_ns = dir->i_sb->s_user_ns; |
2747 | if (!kuid_has_mapping(s_user_ns, current_fsuid()) || | |
2748 | !kgid_has_mapping(s_user_ns, current_fsgid())) | |
2749 | return -EOVERFLOW; | |
f419a2e3 | 2750 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2751 | } |
2752 | ||
1da177e4 LT |
2753 | /* |
2754 | * p1 and p2 should be directories on the same fs. | |
2755 | */ | |
2756 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
2757 | { | |
2758 | struct dentry *p; | |
2759 | ||
2760 | if (p1 == p2) { | |
5955102c | 2761 | inode_lock_nested(p1->d_inode, I_MUTEX_PARENT); |
1da177e4 LT |
2762 | return NULL; |
2763 | } | |
2764 | ||
fc64005c | 2765 | mutex_lock(&p1->d_sb->s_vfs_rename_mutex); |
1da177e4 | 2766 | |
e2761a11 OH |
2767 | p = d_ancestor(p2, p1); |
2768 | if (p) { | |
5955102c AV |
2769 | inode_lock_nested(p2->d_inode, I_MUTEX_PARENT); |
2770 | inode_lock_nested(p1->d_inode, I_MUTEX_CHILD); | |
e2761a11 | 2771 | return p; |
1da177e4 LT |
2772 | } |
2773 | ||
e2761a11 OH |
2774 | p = d_ancestor(p1, p2); |
2775 | if (p) { | |
5955102c AV |
2776 | inode_lock_nested(p1->d_inode, I_MUTEX_PARENT); |
2777 | inode_lock_nested(p2->d_inode, I_MUTEX_CHILD); | |
e2761a11 | 2778 | return p; |
1da177e4 LT |
2779 | } |
2780 | ||
5955102c AV |
2781 | inode_lock_nested(p1->d_inode, I_MUTEX_PARENT); |
2782 | inode_lock_nested(p2->d_inode, I_MUTEX_PARENT2); | |
1da177e4 LT |
2783 | return NULL; |
2784 | } | |
4d359507 | 2785 | EXPORT_SYMBOL(lock_rename); |
1da177e4 LT |
2786 | |
2787 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
2788 | { | |
5955102c | 2789 | inode_unlock(p1->d_inode); |
1da177e4 | 2790 | if (p1 != p2) { |
5955102c | 2791 | inode_unlock(p2->d_inode); |
fc64005c | 2792 | mutex_unlock(&p1->d_sb->s_vfs_rename_mutex); |
1da177e4 LT |
2793 | } |
2794 | } | |
4d359507 | 2795 | EXPORT_SYMBOL(unlock_rename); |
1da177e4 | 2796 | |
4acdaf27 | 2797 | int vfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, |
312b63fb | 2798 | bool want_excl) |
1da177e4 | 2799 | { |
a95164d9 | 2800 | int error = may_create(dir, dentry); |
1da177e4 LT |
2801 | if (error) |
2802 | return error; | |
2803 | ||
acfa4380 | 2804 | if (!dir->i_op->create) |
1da177e4 LT |
2805 | return -EACCES; /* shouldn't it be ENOSYS? */ |
2806 | mode &= S_IALLUGO; | |
2807 | mode |= S_IFREG; | |
2808 | error = security_inode_create(dir, dentry, mode); | |
2809 | if (error) | |
2810 | return error; | |
312b63fb | 2811 | error = dir->i_op->create(dir, dentry, mode, want_excl); |
a74574aa | 2812 | if (!error) |
f38aa942 | 2813 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2814 | return error; |
2815 | } | |
4d359507 | 2816 | EXPORT_SYMBOL(vfs_create); |
1da177e4 | 2817 | |
8e6c848e AV |
2818 | int vfs_mkobj(struct dentry *dentry, umode_t mode, |
2819 | int (*f)(struct dentry *, umode_t, void *), | |
2820 | void *arg) | |
2821 | { | |
2822 | struct inode *dir = dentry->d_parent->d_inode; | |
2823 | int error = may_create(dir, dentry); | |
2824 | if (error) | |
2825 | return error; | |
2826 | ||
2827 | mode &= S_IALLUGO; | |
2828 | mode |= S_IFREG; | |
2829 | error = security_inode_create(dir, dentry, mode); | |
2830 | if (error) | |
2831 | return error; | |
2832 | error = f(dentry, mode, arg); | |
2833 | if (!error) | |
2834 | fsnotify_create(dir, dentry); | |
2835 | return error; | |
2836 | } | |
2837 | EXPORT_SYMBOL(vfs_mkobj); | |
2838 | ||
a2982cc9 EB |
2839 | bool may_open_dev(const struct path *path) |
2840 | { | |
2841 | return !(path->mnt->mnt_flags & MNT_NODEV) && | |
2842 | !(path->mnt->mnt_sb->s_iflags & SB_I_NODEV); | |
2843 | } | |
2844 | ||
f0bb5aaf | 2845 | static int may_open(const struct path *path, int acc_mode, int flag) |
1da177e4 | 2846 | { |
3fb64190 | 2847 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
2848 | struct inode *inode = dentry->d_inode; |
2849 | int error; | |
2850 | ||
2851 | if (!inode) | |
2852 | return -ENOENT; | |
2853 | ||
c8fe8f30 CH |
2854 | switch (inode->i_mode & S_IFMT) { |
2855 | case S_IFLNK: | |
1da177e4 | 2856 | return -ELOOP; |
c8fe8f30 | 2857 | case S_IFDIR: |
fc4177be | 2858 | if (acc_mode & MAY_WRITE) |
c8fe8f30 | 2859 | return -EISDIR; |
fc4177be KC |
2860 | if (acc_mode & MAY_EXEC) |
2861 | return -EACCES; | |
c8fe8f30 CH |
2862 | break; |
2863 | case S_IFBLK: | |
2864 | case S_IFCHR: | |
a2982cc9 | 2865 | if (!may_open_dev(path)) |
1da177e4 | 2866 | return -EACCES; |
633fb6ac | 2867 | fallthrough; |
c8fe8f30 CH |
2868 | case S_IFIFO: |
2869 | case S_IFSOCK: | |
633fb6ac KC |
2870 | if (acc_mode & MAY_EXEC) |
2871 | return -EACCES; | |
1da177e4 | 2872 | flag &= ~O_TRUNC; |
c8fe8f30 | 2873 | break; |
0fd338b2 KC |
2874 | case S_IFREG: |
2875 | if ((acc_mode & MAY_EXEC) && path_noexec(path)) | |
2876 | return -EACCES; | |
2877 | break; | |
4a3fd211 | 2878 | } |
b41572e9 | 2879 | |
62fb4a15 | 2880 | error = inode_permission(inode, MAY_OPEN | acc_mode); |
b41572e9 DH |
2881 | if (error) |
2882 | return error; | |
6146f0d5 | 2883 | |
1da177e4 LT |
2884 | /* |
2885 | * An append-only file must be opened in append mode for writing. | |
2886 | */ | |
2887 | if (IS_APPEND(inode)) { | |
8737c930 | 2888 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 2889 | return -EPERM; |
1da177e4 | 2890 | if (flag & O_TRUNC) |
7715b521 | 2891 | return -EPERM; |
1da177e4 LT |
2892 | } |
2893 | ||
2894 | /* O_NOATIME can only be set by the owner or superuser */ | |
2e149670 | 2895 | if (flag & O_NOATIME && !inode_owner_or_capable(inode)) |
7715b521 | 2896 | return -EPERM; |
1da177e4 | 2897 | |
f3c7691e | 2898 | return 0; |
7715b521 | 2899 | } |
1da177e4 | 2900 | |
e1181ee6 | 2901 | static int handle_truncate(struct file *filp) |
7715b521 | 2902 | { |
f0bb5aaf | 2903 | const struct path *path = &filp->f_path; |
7715b521 AV |
2904 | struct inode *inode = path->dentry->d_inode; |
2905 | int error = get_write_access(inode); | |
2906 | if (error) | |
2907 | return error; | |
2908 | /* | |
2909 | * Refuse to truncate files with mandatory locks held on them. | |
2910 | */ | |
d7a06983 | 2911 | error = locks_verify_locked(filp); |
7715b521 | 2912 | if (!error) |
ea0d3ab2 | 2913 | error = security_path_truncate(path); |
7715b521 AV |
2914 | if (!error) { |
2915 | error = do_truncate(path->dentry, 0, | |
2916 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
e1181ee6 | 2917 | filp); |
7715b521 AV |
2918 | } |
2919 | put_write_access(inode); | |
acd0c935 | 2920 | return error; |
1da177e4 LT |
2921 | } |
2922 | ||
d57999e1 DH |
2923 | static inline int open_to_namei_flags(int flag) |
2924 | { | |
8a5e929d AV |
2925 | if ((flag & O_ACCMODE) == 3) |
2926 | flag--; | |
d57999e1 DH |
2927 | return flag; |
2928 | } | |
2929 | ||
d3607752 | 2930 | static int may_o_create(const struct path *dir, struct dentry *dentry, umode_t mode) |
d18e9008 | 2931 | { |
1328c727 | 2932 | struct user_namespace *s_user_ns; |
d18e9008 MS |
2933 | int error = security_path_mknod(dir, dentry, mode, 0); |
2934 | if (error) | |
2935 | return error; | |
2936 | ||
1328c727 SF |
2937 | s_user_ns = dir->dentry->d_sb->s_user_ns; |
2938 | if (!kuid_has_mapping(s_user_ns, current_fsuid()) || | |
2939 | !kgid_has_mapping(s_user_ns, current_fsgid())) | |
2940 | return -EOVERFLOW; | |
2941 | ||
d18e9008 MS |
2942 | error = inode_permission(dir->dentry->d_inode, MAY_WRITE | MAY_EXEC); |
2943 | if (error) | |
2944 | return error; | |
2945 | ||
2946 | return security_inode_create(dir->dentry->d_inode, dentry, mode); | |
2947 | } | |
2948 | ||
1acf0af9 DH |
2949 | /* |
2950 | * Attempt to atomically look up, create and open a file from a negative | |
2951 | * dentry. | |
2952 | * | |
2953 | * Returns 0 if successful. The file will have been created and attached to | |
2954 | * @file by the filesystem calling finish_open(). | |
2955 | * | |
00a07c15 AV |
2956 | * If the file was looked up only or didn't need creating, FMODE_OPENED won't |
2957 | * be set. The caller will need to perform the open themselves. @path will | |
2958 | * have been updated to point to the new dentry. This may be negative. | |
1acf0af9 DH |
2959 | * |
2960 | * Returns an error code otherwise. | |
2961 | */ | |
239eb983 AV |
2962 | static struct dentry *atomic_open(struct nameidata *nd, struct dentry *dentry, |
2963 | struct file *file, | |
239eb983 | 2964 | int open_flag, umode_t mode) |
d18e9008 | 2965 | { |
384f26e2 | 2966 | struct dentry *const DENTRY_NOT_SET = (void *) -1UL; |
d18e9008 | 2967 | struct inode *dir = nd->path.dentry->d_inode; |
d18e9008 | 2968 | int error; |
d18e9008 | 2969 | |
d18e9008 MS |
2970 | if (nd->flags & LOOKUP_DIRECTORY) |
2971 | open_flag |= O_DIRECTORY; | |
2972 | ||
30d90494 AV |
2973 | file->f_path.dentry = DENTRY_NOT_SET; |
2974 | file->f_path.mnt = nd->path.mnt; | |
0fb1ea09 | 2975 | error = dir->i_op->atomic_open(dir, dentry, file, |
44907d79 | 2976 | open_to_namei_flags(open_flag), mode); |
6fbd0714 | 2977 | d_lookup_done(dentry); |
384f26e2 | 2978 | if (!error) { |
64e1ac4d | 2979 | if (file->f_mode & FMODE_OPENED) { |
6fb968cd AV |
2980 | if (unlikely(dentry != file->f_path.dentry)) { |
2981 | dput(dentry); | |
2982 | dentry = dget(file->f_path.dentry); | |
2983 | } | |
64e1ac4d | 2984 | } else if (WARN_ON(file->f_path.dentry == DENTRY_NOT_SET)) { |
2675a4eb | 2985 | error = -EIO; |
03da633a | 2986 | } else { |
384f26e2 AV |
2987 | if (file->f_path.dentry) { |
2988 | dput(dentry); | |
2989 | dentry = file->f_path.dentry; | |
03da633a | 2990 | } |
239eb983 | 2991 | if (unlikely(d_is_negative(dentry))) |
a01e718f | 2992 | error = -ENOENT; |
62b2ce96 | 2993 | } |
d18e9008 | 2994 | } |
239eb983 AV |
2995 | if (error) { |
2996 | dput(dentry); | |
2997 | dentry = ERR_PTR(error); | |
2998 | } | |
2999 | return dentry; | |
d18e9008 MS |
3000 | } |
3001 | ||
d58ffd35 | 3002 | /* |
1acf0af9 | 3003 | * Look up and maybe create and open the last component. |
d58ffd35 | 3004 | * |
00a07c15 | 3005 | * Must be called with parent locked (exclusive in O_CREAT case). |
1acf0af9 | 3006 | * |
00a07c15 AV |
3007 | * Returns 0 on success, that is, if |
3008 | * the file was successfully atomically created (if necessary) and opened, or | |
3009 | * the file was not completely opened at this time, though lookups and | |
3010 | * creations were performed. | |
3011 | * These case are distinguished by presence of FMODE_OPENED on file->f_mode. | |
3012 | * In the latter case dentry returned in @path might be negative if O_CREAT | |
3013 | * hadn't been specified. | |
1acf0af9 | 3014 | * |
00a07c15 | 3015 | * An error code is returned on failure. |
d58ffd35 | 3016 | */ |
da5ebf5a AV |
3017 | static struct dentry *lookup_open(struct nameidata *nd, struct file *file, |
3018 | const struct open_flags *op, | |
3019 | bool got_write) | |
d58ffd35 MS |
3020 | { |
3021 | struct dentry *dir = nd->path.dentry; | |
54ef4872 | 3022 | struct inode *dir_inode = dir->d_inode; |
1643b43f | 3023 | int open_flag = op->open_flag; |
d58ffd35 | 3024 | struct dentry *dentry; |
1643b43f | 3025 | int error, create_error = 0; |
1643b43f | 3026 | umode_t mode = op->mode; |
6fbd0714 | 3027 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
d58ffd35 | 3028 | |
ce8644fc | 3029 | if (unlikely(IS_DEADDIR(dir_inode))) |
da5ebf5a | 3030 | return ERR_PTR(-ENOENT); |
d58ffd35 | 3031 | |
73a09dd9 | 3032 | file->f_mode &= ~FMODE_CREATED; |
6fbd0714 AV |
3033 | dentry = d_lookup(dir, &nd->last); |
3034 | for (;;) { | |
3035 | if (!dentry) { | |
3036 | dentry = d_alloc_parallel(dir, &nd->last, &wq); | |
3037 | if (IS_ERR(dentry)) | |
da5ebf5a | 3038 | return dentry; |
6fbd0714 AV |
3039 | } |
3040 | if (d_in_lookup(dentry)) | |
3041 | break; | |
d58ffd35 | 3042 | |
6fbd0714 AV |
3043 | error = d_revalidate(dentry, nd->flags); |
3044 | if (likely(error > 0)) | |
3045 | break; | |
3046 | if (error) | |
3047 | goto out_dput; | |
3048 | d_invalidate(dentry); | |
3049 | dput(dentry); | |
3050 | dentry = NULL; | |
3051 | } | |
3052 | if (dentry->d_inode) { | |
6c51e513 | 3053 | /* Cached positive dentry: will open in f_op->open */ |
da5ebf5a | 3054 | return dentry; |
6c51e513 | 3055 | } |
d18e9008 | 3056 | |
1643b43f AV |
3057 | /* |
3058 | * Checking write permission is tricky, bacuse we don't know if we are | |
3059 | * going to actually need it: O_CREAT opens should work as long as the | |
3060 | * file exists. But checking existence breaks atomicity. The trick is | |
3061 | * to check access and if not granted clear O_CREAT from the flags. | |
3062 | * | |
3063 | * Another problem is returing the "right" error value (e.g. for an | |
3064 | * O_EXCL open we want to return EEXIST not EROFS). | |
3065 | */ | |
99a4a90c AV |
3066 | if (unlikely(!got_write)) |
3067 | open_flag &= ~O_TRUNC; | |
1643b43f | 3068 | if (open_flag & O_CREAT) { |
99a4a90c AV |
3069 | if (open_flag & O_EXCL) |
3070 | open_flag &= ~O_TRUNC; | |
1643b43f AV |
3071 | if (!IS_POSIXACL(dir->d_inode)) |
3072 | mode &= ~current_umask(); | |
99a4a90c | 3073 | if (likely(got_write)) |
1643b43f | 3074 | create_error = may_o_create(&nd->path, dentry, mode); |
99a4a90c AV |
3075 | else |
3076 | create_error = -EROFS; | |
d18e9008 | 3077 | } |
99a4a90c AV |
3078 | if (create_error) |
3079 | open_flag &= ~O_CREAT; | |
6ac08709 | 3080 | if (dir_inode->i_op->atomic_open) { |
d489cf9a | 3081 | dentry = atomic_open(nd, dentry, file, open_flag, mode); |
da5ebf5a AV |
3082 | if (unlikely(create_error) && dentry == ERR_PTR(-ENOENT)) |
3083 | dentry = ERR_PTR(create_error); | |
3084 | return dentry; | |
d18e9008 | 3085 | } |
54ef4872 | 3086 | |
6fbd0714 | 3087 | if (d_in_lookup(dentry)) { |
12fa5e24 AV |
3088 | struct dentry *res = dir_inode->i_op->lookup(dir_inode, dentry, |
3089 | nd->flags); | |
6fbd0714 | 3090 | d_lookup_done(dentry); |
12fa5e24 AV |
3091 | if (unlikely(res)) { |
3092 | if (IS_ERR(res)) { | |
3093 | error = PTR_ERR(res); | |
3094 | goto out_dput; | |
3095 | } | |
3096 | dput(dentry); | |
3097 | dentry = res; | |
3098 | } | |
54ef4872 MS |
3099 | } |
3100 | ||
d58ffd35 | 3101 | /* Negative dentry, just create the file */ |
1643b43f | 3102 | if (!dentry->d_inode && (open_flag & O_CREAT)) { |
73a09dd9 | 3103 | file->f_mode |= FMODE_CREATED; |
ce8644fc | 3104 | audit_inode_child(dir_inode, dentry, AUDIT_TYPE_CHILD_CREATE); |
ce8644fc AV |
3105 | if (!dir_inode->i_op->create) { |
3106 | error = -EACCES; | |
d58ffd35 | 3107 | goto out_dput; |
ce8644fc AV |
3108 | } |
3109 | error = dir_inode->i_op->create(dir_inode, dentry, mode, | |
1643b43f | 3110 | open_flag & O_EXCL); |
d58ffd35 MS |
3111 | if (error) |
3112 | goto out_dput; | |
3113 | } | |
1643b43f AV |
3114 | if (unlikely(create_error) && !dentry->d_inode) { |
3115 | error = create_error; | |
3116 | goto out_dput; | |
d58ffd35 | 3117 | } |
da5ebf5a | 3118 | return dentry; |
d58ffd35 MS |
3119 | |
3120 | out_dput: | |
3121 | dput(dentry); | |
da5ebf5a | 3122 | return ERR_PTR(error); |
d58ffd35 MS |
3123 | } |
3124 | ||
c981a482 | 3125 | static const char *open_last_lookups(struct nameidata *nd, |
3ec2eef1 | 3126 | struct file *file, const struct open_flags *op) |
fb1cc555 | 3127 | { |
a1e28038 | 3128 | struct dentry *dir = nd->path.dentry; |
ca344a89 | 3129 | int open_flag = op->open_flag; |
64894cf8 | 3130 | bool got_write = false; |
254cf582 | 3131 | unsigned seq; |
a1eb3315 | 3132 | struct inode *inode; |
da5ebf5a | 3133 | struct dentry *dentry; |
b0417d2c | 3134 | const char *res; |
16c2cd71 | 3135 | int error; |
1f36f774 | 3136 | |
c3e380b0 AV |
3137 | nd->flags |= op->intent; |
3138 | ||
bc77daa7 | 3139 | if (nd->last_type != LAST_NORM) { |
56676ec3 AV |
3140 | if (nd->depth) |
3141 | put_link(nd); | |
ff326a32 | 3142 | return handle_dots(nd, nd->last_type); |
1f36f774 | 3143 | } |
67ee3ad2 | 3144 | |
ca344a89 | 3145 | if (!(open_flag & O_CREAT)) { |
fe2d35ff AV |
3146 | if (nd->last.name[nd->last.len]) |
3147 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
3148 | /* we _can_ be in RCU mode here */ | |
20e34357 AV |
3149 | dentry = lookup_fast(nd, &inode, &seq); |
3150 | if (IS_ERR(dentry)) | |
1ccac622 | 3151 | return ERR_CAST(dentry); |
20e34357 | 3152 | if (likely(dentry)) |
71574865 MS |
3153 | goto finish_lookup; |
3154 | ||
6583fe22 | 3155 | BUG_ON(nd->flags & LOOKUP_RCU); |
b6183df7 MS |
3156 | } else { |
3157 | /* create side of things */ | |
72287417 AV |
3158 | if (nd->flags & LOOKUP_RCU) { |
3159 | error = unlazy_walk(nd); | |
3160 | if (unlikely(error)) | |
3161 | return ERR_PTR(error); | |
3162 | } | |
c9b07eab | 3163 | audit_inode(nd->name, dir, AUDIT_INODE_PARENT); |
b6183df7 | 3164 | /* trailing slashes? */ |
deb106c6 | 3165 | if (unlikely(nd->last.name[nd->last.len])) |
1ccac622 | 3166 | return ERR_PTR(-EISDIR); |
b6183df7 | 3167 | } |
a2c36b45 | 3168 | |
9cf843e3 | 3169 | if (open_flag & (O_CREAT | O_TRUNC | O_WRONLY | O_RDWR)) { |
64894cf8 AV |
3170 | error = mnt_want_write(nd->path.mnt); |
3171 | if (!error) | |
3172 | got_write = true; | |
3173 | /* | |
3174 | * do _not_ fail yet - we might not need that or fail with | |
3175 | * a different error; let lookup_open() decide; we'll be | |
3176 | * dropping this one anyway. | |
3177 | */ | |
3178 | } | |
9cf843e3 AV |
3179 | if (open_flag & O_CREAT) |
3180 | inode_lock(dir->d_inode); | |
3181 | else | |
3182 | inode_lock_shared(dir->d_inode); | |
da5ebf5a | 3183 | dentry = lookup_open(nd, file, op, got_write); |
f7bb959d AV |
3184 | if (!IS_ERR(dentry) && (file->f_mode & FMODE_CREATED)) |
3185 | fsnotify_create(dir->d_inode, dentry); | |
9cf843e3 AV |
3186 | if (open_flag & O_CREAT) |
3187 | inode_unlock(dir->d_inode); | |
3188 | else | |
3189 | inode_unlock_shared(dir->d_inode); | |
a1e28038 | 3190 | |
c981a482 | 3191 | if (got_write) |
59e96e65 | 3192 | mnt_drop_write(nd->path.mnt); |
d18e9008 | 3193 | |
59e96e65 AV |
3194 | if (IS_ERR(dentry)) |
3195 | return ERR_CAST(dentry); | |
3196 | ||
973d4b73 | 3197 | if (file->f_mode & (FMODE_OPENED | FMODE_CREATED)) { |
e73cabff AV |
3198 | dput(nd->path.dentry); |
3199 | nd->path.dentry = dentry; | |
c981a482 | 3200 | return NULL; |
fb1cc555 AV |
3201 | } |
3202 | ||
20e34357 | 3203 | finish_lookup: |
56676ec3 AV |
3204 | if (nd->depth) |
3205 | put_link(nd); | |
8c4efe22 | 3206 | res = step_into(nd, WALK_TRAILING, dentry, inode, seq); |
ff326a32 | 3207 | if (unlikely(res)) |
b0417d2c | 3208 | nd->flags &= ~(LOOKUP_OPEN|LOOKUP_CREATE|LOOKUP_EXCL); |
ff326a32 | 3209 | return res; |
c981a482 AV |
3210 | } |
3211 | ||
3212 | /* | |
3213 | * Handle the last step of open() | |
3214 | */ | |
c5971b8c | 3215 | static int do_open(struct nameidata *nd, |
c981a482 AV |
3216 | struct file *file, const struct open_flags *op) |
3217 | { | |
c981a482 AV |
3218 | int open_flag = op->open_flag; |
3219 | bool do_truncate; | |
3220 | int acc_mode; | |
c981a482 AV |
3221 | int error; |
3222 | ||
ff326a32 AV |
3223 | if (!(file->f_mode & (FMODE_OPENED | FMODE_CREATED))) { |
3224 | error = complete_walk(nd); | |
3225 | if (error) | |
3226 | return error; | |
3227 | } | |
973d4b73 AV |
3228 | if (!(file->f_mode & FMODE_CREATED)) |
3229 | audit_inode(nd->name, nd->path.dentry, 0); | |
30aba665 | 3230 | if (open_flag & O_CREAT) { |
b94e0b32 AV |
3231 | if ((open_flag & O_EXCL) && !(file->f_mode & FMODE_CREATED)) |
3232 | return -EEXIST; | |
30aba665 | 3233 | if (d_is_dir(nd->path.dentry)) |
c5971b8c | 3234 | return -EISDIR; |
0f705953 | 3235 | error = may_create_in_sticky(nd->dir_mode, nd->dir_uid, |
30aba665 SM |
3236 | d_backing_inode(nd->path.dentry)); |
3237 | if (unlikely(error)) | |
c5971b8c | 3238 | return error; |
30aba665 | 3239 | } |
44b1d530 | 3240 | if ((nd->flags & LOOKUP_DIRECTORY) && !d_can_lookup(nd->path.dentry)) |
c5971b8c | 3241 | return -ENOTDIR; |
6c0d46c4 | 3242 | |
8795e7d4 AV |
3243 | do_truncate = false; |
3244 | acc_mode = op->acc_mode; | |
5a2d3edd AV |
3245 | if (file->f_mode & FMODE_CREATED) { |
3246 | /* Don't check for write permission, don't truncate */ | |
3247 | open_flag &= ~O_TRUNC; | |
5a2d3edd | 3248 | acc_mode = 0; |
8795e7d4 | 3249 | } else if (d_is_reg(nd->path.dentry) && open_flag & O_TRUNC) { |
0f9d1a10 AV |
3250 | error = mnt_want_write(nd->path.mnt); |
3251 | if (error) | |
c5971b8c | 3252 | return error; |
8795e7d4 | 3253 | do_truncate = true; |
0f9d1a10 | 3254 | } |
6ac08709 | 3255 | error = may_open(&nd->path, acc_mode, open_flag); |
8795e7d4 | 3256 | if (!error && !(file->f_mode & FMODE_OPENED)) |
3ad5615a | 3257 | error = vfs_open(&nd->path, file); |
8795e7d4 AV |
3258 | if (!error) |
3259 | error = ima_file_check(file, op->acc_mode); | |
3260 | if (!error && do_truncate) | |
2675a4eb | 3261 | error = handle_truncate(file); |
c80567c8 AV |
3262 | if (unlikely(error > 0)) { |
3263 | WARN_ON(1); | |
3264 | error = -EINVAL; | |
3265 | } | |
8795e7d4 | 3266 | if (do_truncate) |
0f9d1a10 | 3267 | mnt_drop_write(nd->path.mnt); |
c5971b8c | 3268 | return error; |
fb1cc555 AV |
3269 | } |
3270 | ||
af7bd4dc AG |
3271 | struct dentry *vfs_tmpfile(struct dentry *dentry, umode_t mode, int open_flag) |
3272 | { | |
af7bd4dc AG |
3273 | struct dentry *child = NULL; |
3274 | struct inode *dir = dentry->d_inode; | |
3275 | struct inode *inode; | |
3276 | int error; | |
3277 | ||
3278 | /* we want directory to be writable */ | |
3279 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); | |
3280 | if (error) | |
3281 | goto out_err; | |
3282 | error = -EOPNOTSUPP; | |
3283 | if (!dir->i_op->tmpfile) | |
3284 | goto out_err; | |
3285 | error = -ENOMEM; | |
cdf01226 | 3286 | child = d_alloc(dentry, &slash_name); |
af7bd4dc AG |
3287 | if (unlikely(!child)) |
3288 | goto out_err; | |
3289 | error = dir->i_op->tmpfile(dir, child, mode); | |
3290 | if (error) | |
3291 | goto out_err; | |
3292 | error = -ENOENT; | |
3293 | inode = child->d_inode; | |
3294 | if (unlikely(!inode)) | |
3295 | goto out_err; | |
3296 | if (!(open_flag & O_EXCL)) { | |
3297 | spin_lock(&inode->i_lock); | |
3298 | inode->i_state |= I_LINKABLE; | |
3299 | spin_unlock(&inode->i_lock); | |
3300 | } | |
fdb2410f | 3301 | ima_post_create_tmpfile(inode); |
af7bd4dc AG |
3302 | return child; |
3303 | ||
3304 | out_err: | |
3305 | dput(child); | |
3306 | return ERR_PTR(error); | |
3307 | } | |
3308 | EXPORT_SYMBOL(vfs_tmpfile); | |
3309 | ||
c8a53ee5 | 3310 | static int do_tmpfile(struct nameidata *nd, unsigned flags, |
60545d0d | 3311 | const struct open_flags *op, |
3ec2eef1 | 3312 | struct file *file) |
60545d0d | 3313 | { |
625b6d10 | 3314 | struct dentry *child; |
625b6d10 | 3315 | struct path path; |
c8a53ee5 | 3316 | int error = path_lookupat(nd, flags | LOOKUP_DIRECTORY, &path); |
60545d0d AV |
3317 | if (unlikely(error)) |
3318 | return error; | |
625b6d10 | 3319 | error = mnt_want_write(path.mnt); |
60545d0d AV |
3320 | if (unlikely(error)) |
3321 | goto out; | |
af7bd4dc AG |
3322 | child = vfs_tmpfile(path.dentry, op->mode, op->open_flag); |
3323 | error = PTR_ERR(child); | |
684e73be | 3324 | if (IS_ERR(child)) |
60545d0d | 3325 | goto out2; |
625b6d10 AV |
3326 | dput(path.dentry); |
3327 | path.dentry = child; | |
c8a53ee5 | 3328 | audit_inode(nd->name, child, 0); |
69a91c23 | 3329 | /* Don't check for other permissions, the inode was just created */ |
62fb4a15 | 3330 | error = may_open(&path, 0, op->open_flag); |
60545d0d AV |
3331 | if (error) |
3332 | goto out2; | |
625b6d10 | 3333 | file->f_path.mnt = path.mnt; |
be12af3e | 3334 | error = finish_open(file, child, NULL); |
60545d0d | 3335 | out2: |
625b6d10 | 3336 | mnt_drop_write(path.mnt); |
60545d0d | 3337 | out: |
625b6d10 | 3338 | path_put(&path); |
60545d0d AV |
3339 | return error; |
3340 | } | |
3341 | ||
6ac08709 AV |
3342 | static int do_o_path(struct nameidata *nd, unsigned flags, struct file *file) |
3343 | { | |
3344 | struct path path; | |
3345 | int error = path_lookupat(nd, flags, &path); | |
3346 | if (!error) { | |
3347 | audit_inode(nd->name, path.dentry, 0); | |
ae2bb293 | 3348 | error = vfs_open(&path, file); |
6ac08709 AV |
3349 | path_put(&path); |
3350 | } | |
3351 | return error; | |
3352 | } | |
3353 | ||
c8a53ee5 AV |
3354 | static struct file *path_openat(struct nameidata *nd, |
3355 | const struct open_flags *op, unsigned flags) | |
1da177e4 | 3356 | { |
30d90494 | 3357 | struct file *file; |
13aab428 | 3358 | int error; |
31e6b01f | 3359 | |
ea73ea72 | 3360 | file = alloc_empty_file(op->open_flag, current_cred()); |
1afc99be AV |
3361 | if (IS_ERR(file)) |
3362 | return file; | |
31e6b01f | 3363 | |
bb458c64 | 3364 | if (unlikely(file->f_flags & __O_TMPFILE)) { |
3ec2eef1 | 3365 | error = do_tmpfile(nd, flags, op, file); |
5f336e72 | 3366 | } else if (unlikely(file->f_flags & O_PATH)) { |
6ac08709 | 3367 | error = do_o_path(nd, flags, file); |
5f336e72 AV |
3368 | } else { |
3369 | const char *s = path_init(nd, flags); | |
3370 | while (!(error = link_path_walk(s, nd)) && | |
c5971b8c | 3371 | (s = open_last_lookups(nd, file, op)) != NULL) |
1ccac622 | 3372 | ; |
c5971b8c AV |
3373 | if (!error) |
3374 | error = do_open(nd, file, op); | |
5f336e72 | 3375 | terminate_walk(nd); |
806b681c | 3376 | } |
7c1c01ec | 3377 | if (likely(!error)) { |
aad888f8 | 3378 | if (likely(file->f_mode & FMODE_OPENED)) |
7c1c01ec AV |
3379 | return file; |
3380 | WARN_ON(1); | |
3381 | error = -EINVAL; | |
16b1c1cd | 3382 | } |
7c1c01ec AV |
3383 | fput(file); |
3384 | if (error == -EOPENSTALE) { | |
3385 | if (flags & LOOKUP_RCU) | |
3386 | error = -ECHILD; | |
3387 | else | |
3388 | error = -ESTALE; | |
2675a4eb | 3389 | } |
7c1c01ec | 3390 | return ERR_PTR(error); |
1da177e4 LT |
3391 | } |
3392 | ||
669abf4e | 3393 | struct file *do_filp_open(int dfd, struct filename *pathname, |
f9652e10 | 3394 | const struct open_flags *op) |
13aab428 | 3395 | { |
9883d185 | 3396 | struct nameidata nd; |
f9652e10 | 3397 | int flags = op->lookup_flags; |
13aab428 AV |
3398 | struct file *filp; |
3399 | ||
9883d185 | 3400 | set_nameidata(&nd, dfd, pathname); |
c8a53ee5 | 3401 | filp = path_openat(&nd, op, flags | LOOKUP_RCU); |
13aab428 | 3402 | if (unlikely(filp == ERR_PTR(-ECHILD))) |
c8a53ee5 | 3403 | filp = path_openat(&nd, op, flags); |
13aab428 | 3404 | if (unlikely(filp == ERR_PTR(-ESTALE))) |
c8a53ee5 | 3405 | filp = path_openat(&nd, op, flags | LOOKUP_REVAL); |
9883d185 | 3406 | restore_nameidata(); |
13aab428 AV |
3407 | return filp; |
3408 | } | |
3409 | ||
73d049a4 | 3410 | struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
f9652e10 | 3411 | const char *name, const struct open_flags *op) |
73d049a4 | 3412 | { |
9883d185 | 3413 | struct nameidata nd; |
73d049a4 | 3414 | struct file *file; |
51689104 | 3415 | struct filename *filename; |
f9652e10 | 3416 | int flags = op->lookup_flags | LOOKUP_ROOT; |
73d049a4 AV |
3417 | |
3418 | nd.root.mnt = mnt; | |
3419 | nd.root.dentry = dentry; | |
3420 | ||
b18825a7 | 3421 | if (d_is_symlink(dentry) && op->intent & LOOKUP_OPEN) |
73d049a4 AV |
3422 | return ERR_PTR(-ELOOP); |
3423 | ||
51689104 | 3424 | filename = getname_kernel(name); |
a1c83681 | 3425 | if (IS_ERR(filename)) |
51689104 PM |
3426 | return ERR_CAST(filename); |
3427 | ||
9883d185 | 3428 | set_nameidata(&nd, -1, filename); |
c8a53ee5 | 3429 | file = path_openat(&nd, op, flags | LOOKUP_RCU); |
73d049a4 | 3430 | if (unlikely(file == ERR_PTR(-ECHILD))) |
c8a53ee5 | 3431 | file = path_openat(&nd, op, flags); |
73d049a4 | 3432 | if (unlikely(file == ERR_PTR(-ESTALE))) |
c8a53ee5 | 3433 | file = path_openat(&nd, op, flags | LOOKUP_REVAL); |
9883d185 | 3434 | restore_nameidata(); |
51689104 | 3435 | putname(filename); |
73d049a4 AV |
3436 | return file; |
3437 | } | |
3438 | ||
fa14a0b8 | 3439 | static struct dentry *filename_create(int dfd, struct filename *name, |
1ac12b4b | 3440 | struct path *path, unsigned int lookup_flags) |
1da177e4 | 3441 | { |
c663e5d8 | 3442 | struct dentry *dentry = ERR_PTR(-EEXIST); |
391172c4 AV |
3443 | struct qstr last; |
3444 | int type; | |
c30dabfe | 3445 | int err2; |
1ac12b4b JL |
3446 | int error; |
3447 | bool is_dir = (lookup_flags & LOOKUP_DIRECTORY); | |
3448 | ||
3449 | /* | |
3450 | * Note that only LOOKUP_REVAL and LOOKUP_DIRECTORY matter here. Any | |
3451 | * other flags passed in are ignored! | |
3452 | */ | |
3453 | lookup_flags &= LOOKUP_REVAL; | |
3454 | ||
5c31b6ce AV |
3455 | name = filename_parentat(dfd, name, lookup_flags, path, &last, &type); |
3456 | if (IS_ERR(name)) | |
3457 | return ERR_CAST(name); | |
1da177e4 | 3458 | |
c663e5d8 CH |
3459 | /* |
3460 | * Yucky last component or no last component at all? | |
3461 | * (foo/., foo/.., /////) | |
3462 | */ | |
5c31b6ce | 3463 | if (unlikely(type != LAST_NORM)) |
ed75e95d | 3464 | goto out; |
c663e5d8 | 3465 | |
c30dabfe | 3466 | /* don't fail immediately if it's r/o, at least try to report other errors */ |
391172c4 | 3467 | err2 = mnt_want_write(path->mnt); |
c663e5d8 CH |
3468 | /* |
3469 | * Do the final lookup. | |
3470 | */ | |
391172c4 | 3471 | lookup_flags |= LOOKUP_CREATE | LOOKUP_EXCL; |
5955102c | 3472 | inode_lock_nested(path->dentry->d_inode, I_MUTEX_PARENT); |
391172c4 | 3473 | dentry = __lookup_hash(&last, path->dentry, lookup_flags); |
1da177e4 | 3474 | if (IS_ERR(dentry)) |
a8104a9f | 3475 | goto unlock; |
c663e5d8 | 3476 | |
a8104a9f | 3477 | error = -EEXIST; |
b18825a7 | 3478 | if (d_is_positive(dentry)) |
a8104a9f | 3479 | goto fail; |
b18825a7 | 3480 | |
c663e5d8 CH |
3481 | /* |
3482 | * Special case - lookup gave negative, but... we had foo/bar/ | |
3483 | * From the vfs_mknod() POV we just have a negative dentry - | |
3484 | * all is fine. Let's be bastards - you had / on the end, you've | |
3485 | * been asking for (non-existent) directory. -ENOENT for you. | |
3486 | */ | |
391172c4 | 3487 | if (unlikely(!is_dir && last.name[last.len])) { |
a8104a9f | 3488 | error = -ENOENT; |
ed75e95d | 3489 | goto fail; |
e9baf6e5 | 3490 | } |
c30dabfe JK |
3491 | if (unlikely(err2)) { |
3492 | error = err2; | |
a8104a9f | 3493 | goto fail; |
c30dabfe | 3494 | } |
181c37b6 | 3495 | putname(name); |
1da177e4 | 3496 | return dentry; |
1da177e4 | 3497 | fail: |
a8104a9f AV |
3498 | dput(dentry); |
3499 | dentry = ERR_PTR(error); | |
3500 | unlock: | |
5955102c | 3501 | inode_unlock(path->dentry->d_inode); |
c30dabfe | 3502 | if (!err2) |
391172c4 | 3503 | mnt_drop_write(path->mnt); |
ed75e95d | 3504 | out: |
391172c4 | 3505 | path_put(path); |
181c37b6 | 3506 | putname(name); |
1da177e4 LT |
3507 | return dentry; |
3508 | } | |
fa14a0b8 AV |
3509 | |
3510 | struct dentry *kern_path_create(int dfd, const char *pathname, | |
3511 | struct path *path, unsigned int lookup_flags) | |
3512 | { | |
181c37b6 AV |
3513 | return filename_create(dfd, getname_kernel(pathname), |
3514 | path, lookup_flags); | |
fa14a0b8 | 3515 | } |
dae6ad8f AV |
3516 | EXPORT_SYMBOL(kern_path_create); |
3517 | ||
921a1650 AV |
3518 | void done_path_create(struct path *path, struct dentry *dentry) |
3519 | { | |
3520 | dput(dentry); | |
5955102c | 3521 | inode_unlock(path->dentry->d_inode); |
a8104a9f | 3522 | mnt_drop_write(path->mnt); |
921a1650 AV |
3523 | path_put(path); |
3524 | } | |
3525 | EXPORT_SYMBOL(done_path_create); | |
3526 | ||
520ae687 | 3527 | inline struct dentry *user_path_create(int dfd, const char __user *pathname, |
1ac12b4b | 3528 | struct path *path, unsigned int lookup_flags) |
dae6ad8f | 3529 | { |
181c37b6 | 3530 | return filename_create(dfd, getname(pathname), path, lookup_flags); |
dae6ad8f AV |
3531 | } |
3532 | EXPORT_SYMBOL(user_path_create); | |
3533 | ||
1a67aafb | 3534 | int vfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev) |
1da177e4 | 3535 | { |
a3c751a5 | 3536 | bool is_whiteout = S_ISCHR(mode) && dev == WHITEOUT_DEV; |
a95164d9 | 3537 | int error = may_create(dir, dentry); |
1da177e4 LT |
3538 | |
3539 | if (error) | |
3540 | return error; | |
3541 | ||
a3c751a5 MS |
3542 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !is_whiteout && |
3543 | !capable(CAP_MKNOD)) | |
1da177e4 LT |
3544 | return -EPERM; |
3545 | ||
acfa4380 | 3546 | if (!dir->i_op->mknod) |
1da177e4 LT |
3547 | return -EPERM; |
3548 | ||
08ce5f16 SH |
3549 | error = devcgroup_inode_mknod(mode, dev); |
3550 | if (error) | |
3551 | return error; | |
3552 | ||
1da177e4 LT |
3553 | error = security_inode_mknod(dir, dentry, mode, dev); |
3554 | if (error) | |
3555 | return error; | |
3556 | ||
1da177e4 | 3557 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 3558 | if (!error) |
f38aa942 | 3559 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3560 | return error; |
3561 | } | |
4d359507 | 3562 | EXPORT_SYMBOL(vfs_mknod); |
1da177e4 | 3563 | |
f69aac00 | 3564 | static int may_mknod(umode_t mode) |
463c3197 DH |
3565 | { |
3566 | switch (mode & S_IFMT) { | |
3567 | case S_IFREG: | |
3568 | case S_IFCHR: | |
3569 | case S_IFBLK: | |
3570 | case S_IFIFO: | |
3571 | case S_IFSOCK: | |
3572 | case 0: /* zero mode translates to S_IFREG */ | |
3573 | return 0; | |
3574 | case S_IFDIR: | |
3575 | return -EPERM; | |
3576 | default: | |
3577 | return -EINVAL; | |
3578 | } | |
3579 | } | |
3580 | ||
5fee64fc | 3581 | static long do_mknodat(int dfd, const char __user *filename, umode_t mode, |
87c4e192 | 3582 | unsigned int dev) |
1da177e4 | 3583 | { |
2ad94ae6 | 3584 | struct dentry *dentry; |
dae6ad8f AV |
3585 | struct path path; |
3586 | int error; | |
972567f1 | 3587 | unsigned int lookup_flags = 0; |
1da177e4 | 3588 | |
8e4bfca1 AV |
3589 | error = may_mknod(mode); |
3590 | if (error) | |
3591 | return error; | |
972567f1 JL |
3592 | retry: |
3593 | dentry = user_path_create(dfd, filename, &path, lookup_flags); | |
dae6ad8f AV |
3594 | if (IS_ERR(dentry)) |
3595 | return PTR_ERR(dentry); | |
2ad94ae6 | 3596 | |
dae6ad8f | 3597 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3598 | mode &= ~current_umask(); |
dae6ad8f | 3599 | error = security_path_mknod(&path, dentry, mode, dev); |
be6d3e56 | 3600 | if (error) |
a8104a9f | 3601 | goto out; |
463c3197 | 3602 | switch (mode & S_IFMT) { |
1da177e4 | 3603 | case 0: case S_IFREG: |
312b63fb | 3604 | error = vfs_create(path.dentry->d_inode,dentry,mode,true); |
05d1a717 MZ |
3605 | if (!error) |
3606 | ima_post_path_mknod(dentry); | |
1da177e4 LT |
3607 | break; |
3608 | case S_IFCHR: case S_IFBLK: | |
dae6ad8f | 3609 | error = vfs_mknod(path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
3610 | new_decode_dev(dev)); |
3611 | break; | |
3612 | case S_IFIFO: case S_IFSOCK: | |
dae6ad8f | 3613 | error = vfs_mknod(path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 3614 | break; |
1da177e4 | 3615 | } |
a8104a9f | 3616 | out: |
921a1650 | 3617 | done_path_create(&path, dentry); |
972567f1 JL |
3618 | if (retry_estale(error, lookup_flags)) { |
3619 | lookup_flags |= LOOKUP_REVAL; | |
3620 | goto retry; | |
3621 | } | |
1da177e4 LT |
3622 | return error; |
3623 | } | |
3624 | ||
87c4e192 DB |
3625 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, umode_t, mode, |
3626 | unsigned int, dev) | |
3627 | { | |
3628 | return do_mknodat(dfd, filename, mode, dev); | |
3629 | } | |
3630 | ||
8208a22b | 3631 | SYSCALL_DEFINE3(mknod, const char __user *, filename, umode_t, mode, unsigned, dev) |
5590ff0d | 3632 | { |
87c4e192 | 3633 | return do_mknodat(AT_FDCWD, filename, mode, dev); |
5590ff0d UD |
3634 | } |
3635 | ||
18bb1db3 | 3636 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) |
1da177e4 | 3637 | { |
a95164d9 | 3638 | int error = may_create(dir, dentry); |
8de52778 | 3639 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
3640 | |
3641 | if (error) | |
3642 | return error; | |
3643 | ||
acfa4380 | 3644 | if (!dir->i_op->mkdir) |
1da177e4 LT |
3645 | return -EPERM; |
3646 | ||
3647 | mode &= (S_IRWXUGO|S_ISVTX); | |
3648 | error = security_inode_mkdir(dir, dentry, mode); | |
3649 | if (error) | |
3650 | return error; | |
3651 | ||
8de52778 AV |
3652 | if (max_links && dir->i_nlink >= max_links) |
3653 | return -EMLINK; | |
3654 | ||
1da177e4 | 3655 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 3656 | if (!error) |
f38aa942 | 3657 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
3658 | return error; |
3659 | } | |
4d359507 | 3660 | EXPORT_SYMBOL(vfs_mkdir); |
1da177e4 | 3661 | |
83ff98c3 | 3662 | static long do_mkdirat(int dfd, const char __user *pathname, umode_t mode) |
1da177e4 | 3663 | { |
6902d925 | 3664 | struct dentry *dentry; |
dae6ad8f AV |
3665 | struct path path; |
3666 | int error; | |
b76d8b82 | 3667 | unsigned int lookup_flags = LOOKUP_DIRECTORY; |
1da177e4 | 3668 | |
b76d8b82 JL |
3669 | retry: |
3670 | dentry = user_path_create(dfd, pathname, &path, lookup_flags); | |
6902d925 | 3671 | if (IS_ERR(dentry)) |
dae6ad8f | 3672 | return PTR_ERR(dentry); |
1da177e4 | 3673 | |
dae6ad8f | 3674 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3675 | mode &= ~current_umask(); |
dae6ad8f | 3676 | error = security_path_mkdir(&path, dentry, mode); |
a8104a9f AV |
3677 | if (!error) |
3678 | error = vfs_mkdir(path.dentry->d_inode, dentry, mode); | |
921a1650 | 3679 | done_path_create(&path, dentry); |
b76d8b82 JL |
3680 | if (retry_estale(error, lookup_flags)) { |
3681 | lookup_flags |= LOOKUP_REVAL; | |
3682 | goto retry; | |
3683 | } | |
1da177e4 LT |
3684 | return error; |
3685 | } | |
3686 | ||
0101db7a DB |
3687 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, umode_t, mode) |
3688 | { | |
3689 | return do_mkdirat(dfd, pathname, mode); | |
3690 | } | |
3691 | ||
a218d0fd | 3692 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, umode_t, mode) |
5590ff0d | 3693 | { |
0101db7a | 3694 | return do_mkdirat(AT_FDCWD, pathname, mode); |
5590ff0d UD |
3695 | } |
3696 | ||
1da177e4 LT |
3697 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) |
3698 | { | |
3699 | int error = may_delete(dir, dentry, 1); | |
3700 | ||
3701 | if (error) | |
3702 | return error; | |
3703 | ||
acfa4380 | 3704 | if (!dir->i_op->rmdir) |
1da177e4 LT |
3705 | return -EPERM; |
3706 | ||
1d2ef590 | 3707 | dget(dentry); |
5955102c | 3708 | inode_lock(dentry->d_inode); |
912dbc15 SW |
3709 | |
3710 | error = -EBUSY; | |
7af1364f | 3711 | if (is_local_mountpoint(dentry)) |
912dbc15 SW |
3712 | goto out; |
3713 | ||
3714 | error = security_inode_rmdir(dir, dentry); | |
3715 | if (error) | |
3716 | goto out; | |
3717 | ||
3718 | error = dir->i_op->rmdir(dir, dentry); | |
3719 | if (error) | |
3720 | goto out; | |
3721 | ||
8767712f | 3722 | shrink_dcache_parent(dentry); |
912dbc15 SW |
3723 | dentry->d_inode->i_flags |= S_DEAD; |
3724 | dont_mount(dentry); | |
8ed936b5 | 3725 | detach_mounts(dentry); |
116b9731 | 3726 | fsnotify_rmdir(dir, dentry); |
912dbc15 SW |
3727 | |
3728 | out: | |
5955102c | 3729 | inode_unlock(dentry->d_inode); |
1d2ef590 | 3730 | dput(dentry); |
912dbc15 | 3731 | if (!error) |
1da177e4 | 3732 | d_delete(dentry); |
1da177e4 LT |
3733 | return error; |
3734 | } | |
4d359507 | 3735 | EXPORT_SYMBOL(vfs_rmdir); |
1da177e4 | 3736 | |
e24ab0ef | 3737 | long do_rmdir(int dfd, struct filename *name) |
1da177e4 LT |
3738 | { |
3739 | int error = 0; | |
1da177e4 | 3740 | struct dentry *dentry; |
f5beed75 AV |
3741 | struct path path; |
3742 | struct qstr last; | |
3743 | int type; | |
c6ee9206 JL |
3744 | unsigned int lookup_flags = 0; |
3745 | retry: | |
e24ab0ef | 3746 | name = filename_parentat(dfd, name, lookup_flags, |
c1d4dd27 | 3747 | &path, &last, &type); |
91a27b2a JL |
3748 | if (IS_ERR(name)) |
3749 | return PTR_ERR(name); | |
1da177e4 | 3750 | |
f5beed75 | 3751 | switch (type) { |
0612d9fb OH |
3752 | case LAST_DOTDOT: |
3753 | error = -ENOTEMPTY; | |
3754 | goto exit1; | |
3755 | case LAST_DOT: | |
3756 | error = -EINVAL; | |
3757 | goto exit1; | |
3758 | case LAST_ROOT: | |
3759 | error = -EBUSY; | |
3760 | goto exit1; | |
1da177e4 | 3761 | } |
0612d9fb | 3762 | |
f5beed75 | 3763 | error = mnt_want_write(path.mnt); |
c30dabfe JK |
3764 | if (error) |
3765 | goto exit1; | |
0612d9fb | 3766 | |
5955102c | 3767 | inode_lock_nested(path.dentry->d_inode, I_MUTEX_PARENT); |
f5beed75 | 3768 | dentry = __lookup_hash(&last, path.dentry, lookup_flags); |
1da177e4 | 3769 | error = PTR_ERR(dentry); |
6902d925 DH |
3770 | if (IS_ERR(dentry)) |
3771 | goto exit2; | |
e6bc45d6 TT |
3772 | if (!dentry->d_inode) { |
3773 | error = -ENOENT; | |
3774 | goto exit3; | |
3775 | } | |
f5beed75 | 3776 | error = security_path_rmdir(&path, dentry); |
be6d3e56 | 3777 | if (error) |
c30dabfe | 3778 | goto exit3; |
f5beed75 | 3779 | error = vfs_rmdir(path.dentry->d_inode, dentry); |
0622753b | 3780 | exit3: |
6902d925 DH |
3781 | dput(dentry); |
3782 | exit2: | |
5955102c | 3783 | inode_unlock(path.dentry->d_inode); |
f5beed75 | 3784 | mnt_drop_write(path.mnt); |
1da177e4 | 3785 | exit1: |
f5beed75 | 3786 | path_put(&path); |
c6ee9206 JL |
3787 | if (retry_estale(error, lookup_flags)) { |
3788 | lookup_flags |= LOOKUP_REVAL; | |
3789 | goto retry; | |
3790 | } | |
24fb33d4 | 3791 | putname(name); |
1da177e4 LT |
3792 | return error; |
3793 | } | |
3794 | ||
3cdad428 | 3795 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d | 3796 | { |
e24ab0ef | 3797 | return do_rmdir(AT_FDCWD, getname(pathname)); |
5590ff0d UD |
3798 | } |
3799 | ||
b21996e3 BF |
3800 | /** |
3801 | * vfs_unlink - unlink a filesystem object | |
3802 | * @dir: parent directory | |
3803 | * @dentry: victim | |
3804 | * @delegated_inode: returns victim inode, if the inode is delegated. | |
3805 | * | |
3806 | * The caller must hold dir->i_mutex. | |
3807 | * | |
3808 | * If vfs_unlink discovers a delegation, it will return -EWOULDBLOCK and | |
3809 | * return a reference to the inode in delegated_inode. The caller | |
3810 | * should then break the delegation on that inode and retry. Because | |
3811 | * breaking a delegation may take a long time, the caller should drop | |
3812 | * dir->i_mutex before doing so. | |
3813 | * | |
3814 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
3815 | * be appropriate for callers that expect the underlying filesystem not | |
3816 | * to be NFS exported. | |
3817 | */ | |
3818 | int vfs_unlink(struct inode *dir, struct dentry *dentry, struct inode **delegated_inode) | |
1da177e4 | 3819 | { |
9accbb97 | 3820 | struct inode *target = dentry->d_inode; |
1da177e4 LT |
3821 | int error = may_delete(dir, dentry, 0); |
3822 | ||
3823 | if (error) | |
3824 | return error; | |
3825 | ||
acfa4380 | 3826 | if (!dir->i_op->unlink) |
1da177e4 LT |
3827 | return -EPERM; |
3828 | ||
5955102c | 3829 | inode_lock(target); |
8ed936b5 | 3830 | if (is_local_mountpoint(dentry)) |
1da177e4 LT |
3831 | error = -EBUSY; |
3832 | else { | |
3833 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 3834 | if (!error) { |
5a14696c BF |
3835 | error = try_break_deleg(target, delegated_inode); |
3836 | if (error) | |
b21996e3 | 3837 | goto out; |
1da177e4 | 3838 | error = dir->i_op->unlink(dir, dentry); |
8ed936b5 | 3839 | if (!error) { |
d83c49f3 | 3840 | dont_mount(dentry); |
8ed936b5 | 3841 | detach_mounts(dentry); |
116b9731 | 3842 | fsnotify_unlink(dir, dentry); |
8ed936b5 | 3843 | } |
bec1052e | 3844 | } |
1da177e4 | 3845 | } |
b21996e3 | 3846 | out: |
5955102c | 3847 | inode_unlock(target); |
1da177e4 LT |
3848 | |
3849 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
3850 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
9accbb97 | 3851 | fsnotify_link_count(target); |
e234f35c | 3852 | d_delete(dentry); |
1da177e4 | 3853 | } |
0eeca283 | 3854 | |
1da177e4 LT |
3855 | return error; |
3856 | } | |
4d359507 | 3857 | EXPORT_SYMBOL(vfs_unlink); |
1da177e4 LT |
3858 | |
3859 | /* | |
3860 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 3861 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
3862 | * writeout happening, and we don't want to prevent access to the directory |
3863 | * while waiting on the I/O. | |
3864 | */ | |
da2f1362 | 3865 | long do_unlinkat(int dfd, struct filename *name) |
1da177e4 | 3866 | { |
2ad94ae6 | 3867 | int error; |
1da177e4 | 3868 | struct dentry *dentry; |
f5beed75 AV |
3869 | struct path path; |
3870 | struct qstr last; | |
3871 | int type; | |
1da177e4 | 3872 | struct inode *inode = NULL; |
b21996e3 | 3873 | struct inode *delegated_inode = NULL; |
5d18f813 JL |
3874 | unsigned int lookup_flags = 0; |
3875 | retry: | |
da2f1362 | 3876 | name = filename_parentat(dfd, name, lookup_flags, &path, &last, &type); |
91a27b2a JL |
3877 | if (IS_ERR(name)) |
3878 | return PTR_ERR(name); | |
2ad94ae6 | 3879 | |
1da177e4 | 3880 | error = -EISDIR; |
f5beed75 | 3881 | if (type != LAST_NORM) |
1da177e4 | 3882 | goto exit1; |
0612d9fb | 3883 | |
f5beed75 | 3884 | error = mnt_want_write(path.mnt); |
c30dabfe JK |
3885 | if (error) |
3886 | goto exit1; | |
b21996e3 | 3887 | retry_deleg: |
5955102c | 3888 | inode_lock_nested(path.dentry->d_inode, I_MUTEX_PARENT); |
f5beed75 | 3889 | dentry = __lookup_hash(&last, path.dentry, lookup_flags); |
1da177e4 LT |
3890 | error = PTR_ERR(dentry); |
3891 | if (!IS_ERR(dentry)) { | |
3892 | /* Why not before? Because we want correct error value */ | |
f5beed75 | 3893 | if (last.name[last.len]) |
50338b88 | 3894 | goto slashes; |
1da177e4 | 3895 | inode = dentry->d_inode; |
b18825a7 | 3896 | if (d_is_negative(dentry)) |
e6bc45d6 TT |
3897 | goto slashes; |
3898 | ihold(inode); | |
f5beed75 | 3899 | error = security_path_unlink(&path, dentry); |
be6d3e56 | 3900 | if (error) |
c30dabfe | 3901 | goto exit2; |
f5beed75 | 3902 | error = vfs_unlink(path.dentry->d_inode, dentry, &delegated_inode); |
c30dabfe | 3903 | exit2: |
1da177e4 LT |
3904 | dput(dentry); |
3905 | } | |
5955102c | 3906 | inode_unlock(path.dentry->d_inode); |
1da177e4 LT |
3907 | if (inode) |
3908 | iput(inode); /* truncate the inode here */ | |
b21996e3 BF |
3909 | inode = NULL; |
3910 | if (delegated_inode) { | |
5a14696c | 3911 | error = break_deleg_wait(&delegated_inode); |
b21996e3 BF |
3912 | if (!error) |
3913 | goto retry_deleg; | |
3914 | } | |
f5beed75 | 3915 | mnt_drop_write(path.mnt); |
1da177e4 | 3916 | exit1: |
f5beed75 | 3917 | path_put(&path); |
5d18f813 JL |
3918 | if (retry_estale(error, lookup_flags)) { |
3919 | lookup_flags |= LOOKUP_REVAL; | |
3920 | inode = NULL; | |
3921 | goto retry; | |
3922 | } | |
da2f1362 | 3923 | putname(name); |
1da177e4 LT |
3924 | return error; |
3925 | ||
3926 | slashes: | |
b18825a7 DH |
3927 | if (d_is_negative(dentry)) |
3928 | error = -ENOENT; | |
44b1d530 | 3929 | else if (d_is_dir(dentry)) |
b18825a7 DH |
3930 | error = -EISDIR; |
3931 | else | |
3932 | error = -ENOTDIR; | |
1da177e4 LT |
3933 | goto exit2; |
3934 | } | |
3935 | ||
2e4d0924 | 3936 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
3937 | { |
3938 | if ((flag & ~AT_REMOVEDIR) != 0) | |
3939 | return -EINVAL; | |
3940 | ||
3941 | if (flag & AT_REMOVEDIR) | |
e24ab0ef | 3942 | return do_rmdir(dfd, getname(pathname)); |
da2f1362 | 3943 | return do_unlinkat(dfd, getname(pathname)); |
5590ff0d UD |
3944 | } |
3945 | ||
3480b257 | 3946 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d | 3947 | { |
da2f1362 | 3948 | return do_unlinkat(AT_FDCWD, getname(pathname)); |
5590ff0d UD |
3949 | } |
3950 | ||
db2e747b | 3951 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 3952 | { |
a95164d9 | 3953 | int error = may_create(dir, dentry); |
1da177e4 LT |
3954 | |
3955 | if (error) | |
3956 | return error; | |
3957 | ||
acfa4380 | 3958 | if (!dir->i_op->symlink) |
1da177e4 LT |
3959 | return -EPERM; |
3960 | ||
3961 | error = security_inode_symlink(dir, dentry, oldname); | |
3962 | if (error) | |
3963 | return error; | |
3964 | ||
1da177e4 | 3965 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 3966 | if (!error) |
f38aa942 | 3967 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3968 | return error; |
3969 | } | |
4d359507 | 3970 | EXPORT_SYMBOL(vfs_symlink); |
1da177e4 | 3971 | |
cd3acb6a | 3972 | static long do_symlinkat(const char __user *oldname, int newdfd, |
b724e846 | 3973 | const char __user *newname) |
1da177e4 | 3974 | { |
2ad94ae6 | 3975 | int error; |
91a27b2a | 3976 | struct filename *from; |
6902d925 | 3977 | struct dentry *dentry; |
dae6ad8f | 3978 | struct path path; |
f46d3567 | 3979 | unsigned int lookup_flags = 0; |
1da177e4 LT |
3980 | |
3981 | from = getname(oldname); | |
2ad94ae6 | 3982 | if (IS_ERR(from)) |
1da177e4 | 3983 | return PTR_ERR(from); |
f46d3567 JL |
3984 | retry: |
3985 | dentry = user_path_create(newdfd, newname, &path, lookup_flags); | |
6902d925 DH |
3986 | error = PTR_ERR(dentry); |
3987 | if (IS_ERR(dentry)) | |
dae6ad8f | 3988 | goto out_putname; |
6902d925 | 3989 | |
91a27b2a | 3990 | error = security_path_symlink(&path, dentry, from->name); |
a8104a9f | 3991 | if (!error) |
91a27b2a | 3992 | error = vfs_symlink(path.dentry->d_inode, dentry, from->name); |
921a1650 | 3993 | done_path_create(&path, dentry); |
f46d3567 JL |
3994 | if (retry_estale(error, lookup_flags)) { |
3995 | lookup_flags |= LOOKUP_REVAL; | |
3996 | goto retry; | |
3997 | } | |
6902d925 | 3998 | out_putname: |
1da177e4 LT |
3999 | putname(from); |
4000 | return error; | |
4001 | } | |
4002 | ||
b724e846 DB |
4003 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
4004 | int, newdfd, const char __user *, newname) | |
4005 | { | |
4006 | return do_symlinkat(oldname, newdfd, newname); | |
4007 | } | |
4008 | ||
3480b257 | 4009 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4010 | { |
b724e846 | 4011 | return do_symlinkat(oldname, AT_FDCWD, newname); |
5590ff0d UD |
4012 | } |
4013 | ||
146a8595 BF |
4014 | /** |
4015 | * vfs_link - create a new link | |
4016 | * @old_dentry: object to be linked | |
4017 | * @dir: new parent | |
4018 | * @new_dentry: where to create the new link | |
4019 | * @delegated_inode: returns inode needing a delegation break | |
4020 | * | |
4021 | * The caller must hold dir->i_mutex | |
4022 | * | |
4023 | * If vfs_link discovers a delegation on the to-be-linked file in need | |
4024 | * of breaking, it will return -EWOULDBLOCK and return a reference to the | |
4025 | * inode in delegated_inode. The caller should then break the delegation | |
4026 | * and retry. Because breaking a delegation may take a long time, the | |
4027 | * caller should drop the i_mutex before doing so. | |
4028 | * | |
4029 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
4030 | * be appropriate for callers that expect the underlying filesystem not | |
4031 | * to be NFS exported. | |
4032 | */ | |
4033 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry, struct inode **delegated_inode) | |
1da177e4 LT |
4034 | { |
4035 | struct inode *inode = old_dentry->d_inode; | |
8de52778 | 4036 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
4037 | int error; |
4038 | ||
4039 | if (!inode) | |
4040 | return -ENOENT; | |
4041 | ||
a95164d9 | 4042 | error = may_create(dir, new_dentry); |
1da177e4 LT |
4043 | if (error) |
4044 | return error; | |
4045 | ||
4046 | if (dir->i_sb != inode->i_sb) | |
4047 | return -EXDEV; | |
4048 | ||
4049 | /* | |
4050 | * A link to an append-only or immutable file cannot be created. | |
4051 | */ | |
4052 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
4053 | return -EPERM; | |
0bd23d09 EB |
4054 | /* |
4055 | * Updating the link count will likely cause i_uid and i_gid to | |
4056 | * be writen back improperly if their true value is unknown to | |
4057 | * the vfs. | |
4058 | */ | |
4059 | if (HAS_UNMAPPED_ID(inode)) | |
4060 | return -EPERM; | |
acfa4380 | 4061 | if (!dir->i_op->link) |
1da177e4 | 4062 | return -EPERM; |
7e79eedb | 4063 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
4064 | return -EPERM; |
4065 | ||
4066 | error = security_inode_link(old_dentry, dir, new_dentry); | |
4067 | if (error) | |
4068 | return error; | |
4069 | ||
5955102c | 4070 | inode_lock(inode); |
aae8a97d | 4071 | /* Make sure we don't allow creating hardlink to an unlinked file */ |
f4e0c30c | 4072 | if (inode->i_nlink == 0 && !(inode->i_state & I_LINKABLE)) |
aae8a97d | 4073 | error = -ENOENT; |
8de52778 AV |
4074 | else if (max_links && inode->i_nlink >= max_links) |
4075 | error = -EMLINK; | |
146a8595 BF |
4076 | else { |
4077 | error = try_break_deleg(inode, delegated_inode); | |
4078 | if (!error) | |
4079 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
4080 | } | |
f4e0c30c AV |
4081 | |
4082 | if (!error && (inode->i_state & I_LINKABLE)) { | |
4083 | spin_lock(&inode->i_lock); | |
4084 | inode->i_state &= ~I_LINKABLE; | |
4085 | spin_unlock(&inode->i_lock); | |
4086 | } | |
5955102c | 4087 | inode_unlock(inode); |
e31e14ec | 4088 | if (!error) |
7e79eedb | 4089 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
4090 | return error; |
4091 | } | |
4d359507 | 4092 | EXPORT_SYMBOL(vfs_link); |
1da177e4 LT |
4093 | |
4094 | /* | |
4095 | * Hardlinks are often used in delicate situations. We avoid | |
4096 | * security-related surprises by not following symlinks on the | |
4097 | * newname. --KAB | |
4098 | * | |
4099 | * We don't follow them on the oldname either to be compatible | |
4100 | * with linux 2.0, and to avoid hard-linking to directories | |
4101 | * and other special files. --ADM | |
4102 | */ | |
812931d6 | 4103 | static int do_linkat(int olddfd, const char __user *oldname, int newdfd, |
46ea89eb | 4104 | const char __user *newname, int flags) |
1da177e4 LT |
4105 | { |
4106 | struct dentry *new_dentry; | |
dae6ad8f | 4107 | struct path old_path, new_path; |
146a8595 | 4108 | struct inode *delegated_inode = NULL; |
11a7b371 | 4109 | int how = 0; |
1da177e4 | 4110 | int error; |
1da177e4 | 4111 | |
11a7b371 | 4112 | if ((flags & ~(AT_SYMLINK_FOLLOW | AT_EMPTY_PATH)) != 0) |
c04030e1 | 4113 | return -EINVAL; |
11a7b371 | 4114 | /* |
f0cc6ffb LT |
4115 | * To use null names we require CAP_DAC_READ_SEARCH |
4116 | * This ensures that not everyone will be able to create | |
4117 | * handlink using the passed filedescriptor. | |
11a7b371 | 4118 | */ |
f0cc6ffb LT |
4119 | if (flags & AT_EMPTY_PATH) { |
4120 | if (!capable(CAP_DAC_READ_SEARCH)) | |
4121 | return -ENOENT; | |
11a7b371 | 4122 | how = LOOKUP_EMPTY; |
f0cc6ffb | 4123 | } |
11a7b371 AK |
4124 | |
4125 | if (flags & AT_SYMLINK_FOLLOW) | |
4126 | how |= LOOKUP_FOLLOW; | |
442e31ca | 4127 | retry: |
11a7b371 | 4128 | error = user_path_at(olddfd, oldname, how, &old_path); |
1da177e4 | 4129 | if (error) |
2ad94ae6 AV |
4130 | return error; |
4131 | ||
442e31ca JL |
4132 | new_dentry = user_path_create(newdfd, newname, &new_path, |
4133 | (how & LOOKUP_REVAL)); | |
1da177e4 | 4134 | error = PTR_ERR(new_dentry); |
6902d925 | 4135 | if (IS_ERR(new_dentry)) |
dae6ad8f AV |
4136 | goto out; |
4137 | ||
4138 | error = -EXDEV; | |
4139 | if (old_path.mnt != new_path.mnt) | |
4140 | goto out_dput; | |
800179c9 KC |
4141 | error = may_linkat(&old_path); |
4142 | if (unlikely(error)) | |
4143 | goto out_dput; | |
dae6ad8f | 4144 | error = security_path_link(old_path.dentry, &new_path, new_dentry); |
be6d3e56 | 4145 | if (error) |
a8104a9f | 4146 | goto out_dput; |
146a8595 | 4147 | error = vfs_link(old_path.dentry, new_path.dentry->d_inode, new_dentry, &delegated_inode); |
75c3f29d | 4148 | out_dput: |
921a1650 | 4149 | done_path_create(&new_path, new_dentry); |
146a8595 BF |
4150 | if (delegated_inode) { |
4151 | error = break_deleg_wait(&delegated_inode); | |
d22e6338 OD |
4152 | if (!error) { |
4153 | path_put(&old_path); | |
146a8595 | 4154 | goto retry; |
d22e6338 | 4155 | } |
146a8595 | 4156 | } |
442e31ca | 4157 | if (retry_estale(error, how)) { |
d22e6338 | 4158 | path_put(&old_path); |
442e31ca JL |
4159 | how |= LOOKUP_REVAL; |
4160 | goto retry; | |
4161 | } | |
1da177e4 | 4162 | out: |
2d8f3038 | 4163 | path_put(&old_path); |
1da177e4 LT |
4164 | |
4165 | return error; | |
4166 | } | |
4167 | ||
46ea89eb DB |
4168 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
4169 | int, newdfd, const char __user *, newname, int, flags) | |
4170 | { | |
4171 | return do_linkat(olddfd, oldname, newdfd, newname, flags); | |
4172 | } | |
4173 | ||
3480b257 | 4174 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4175 | { |
46ea89eb | 4176 | return do_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
4177 | } |
4178 | ||
bc27027a MS |
4179 | /** |
4180 | * vfs_rename - rename a filesystem object | |
4181 | * @old_dir: parent of source | |
4182 | * @old_dentry: source | |
4183 | * @new_dir: parent of destination | |
4184 | * @new_dentry: destination | |
4185 | * @delegated_inode: returns an inode needing a delegation break | |
520c8b16 | 4186 | * @flags: rename flags |
bc27027a MS |
4187 | * |
4188 | * The caller must hold multiple mutexes--see lock_rename()). | |
4189 | * | |
4190 | * If vfs_rename discovers a delegation in need of breaking at either | |
4191 | * the source or destination, it will return -EWOULDBLOCK and return a | |
4192 | * reference to the inode in delegated_inode. The caller should then | |
4193 | * break the delegation and retry. Because breaking a delegation may | |
4194 | * take a long time, the caller should drop all locks before doing | |
4195 | * so. | |
4196 | * | |
4197 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
4198 | * be appropriate for callers that expect the underlying filesystem not | |
4199 | * to be NFS exported. | |
4200 | * | |
1da177e4 LT |
4201 | * The worst of all namespace operations - renaming directory. "Perverted" |
4202 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
4203 | * Problems: | |
0117d427 | 4204 | * |
d03b29a2 | 4205 | * a) we can get into loop creation. |
1da177e4 LT |
4206 | * b) race potential - two innocent renames can create a loop together. |
4207 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 4208 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 | 4209 | * story. |
6cedba89 BF |
4210 | * c) we have to lock _four_ objects - parents and victim (if it exists), |
4211 | * and source (if it is not a directory). | |
1b1dcc1b | 4212 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
4213 | * whether the target exists). Solution: try to be smart with locking |
4214 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 4215 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
4216 | * move will be locked. Thus we can rank directories by the tree |
4217 | * (ancestors first) and rank all non-directories after them. | |
4218 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 4219 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
4220 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
4221 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
4222 | * we'd better make sure that there's no link(2) for them. | |
e4eaac06 | 4223 | * d) conversion from fhandle to dentry may come in the wrong moment - when |
1b1dcc1b | 4224 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 4225 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 4226 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
4227 | * locking]. |
4228 | */ | |
bc27027a MS |
4229 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, |
4230 | struct inode *new_dir, struct dentry *new_dentry, | |
520c8b16 | 4231 | struct inode **delegated_inode, unsigned int flags) |
1da177e4 | 4232 | { |
bc27027a MS |
4233 | int error; |
4234 | bool is_dir = d_is_dir(old_dentry); | |
bc27027a | 4235 | struct inode *source = old_dentry->d_inode; |
9055cba7 | 4236 | struct inode *target = new_dentry->d_inode; |
da1ce067 MS |
4237 | bool new_is_dir = false; |
4238 | unsigned max_links = new_dir->i_sb->s_max_links; | |
49d31c2f | 4239 | struct name_snapshot old_name; |
bc27027a | 4240 | |
8d3e2936 | 4241 | if (source == target) |
bc27027a MS |
4242 | return 0; |
4243 | ||
4244 | error = may_delete(old_dir, old_dentry, is_dir); | |
4245 | if (error) | |
4246 | return error; | |
4247 | ||
da1ce067 | 4248 | if (!target) { |
bc27027a | 4249 | error = may_create(new_dir, new_dentry); |
da1ce067 MS |
4250 | } else { |
4251 | new_is_dir = d_is_dir(new_dentry); | |
4252 | ||
4253 | if (!(flags & RENAME_EXCHANGE)) | |
4254 | error = may_delete(new_dir, new_dentry, is_dir); | |
4255 | else | |
4256 | error = may_delete(new_dir, new_dentry, new_is_dir); | |
4257 | } | |
bc27027a MS |
4258 | if (error) |
4259 | return error; | |
4260 | ||
2773bf00 | 4261 | if (!old_dir->i_op->rename) |
bc27027a | 4262 | return -EPERM; |
1da177e4 LT |
4263 | |
4264 | /* | |
4265 | * If we are going to change the parent - check write permissions, | |
4266 | * we'll need to flip '..'. | |
4267 | */ | |
da1ce067 MS |
4268 | if (new_dir != old_dir) { |
4269 | if (is_dir) { | |
4270 | error = inode_permission(source, MAY_WRITE); | |
4271 | if (error) | |
4272 | return error; | |
4273 | } | |
4274 | if ((flags & RENAME_EXCHANGE) && new_is_dir) { | |
4275 | error = inode_permission(target, MAY_WRITE); | |
4276 | if (error) | |
4277 | return error; | |
4278 | } | |
1da177e4 LT |
4279 | } |
4280 | ||
0b3974eb MS |
4281 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry, |
4282 | flags); | |
1da177e4 LT |
4283 | if (error) |
4284 | return error; | |
4285 | ||
49d31c2f | 4286 | take_dentry_name_snapshot(&old_name, old_dentry); |
1d2ef590 | 4287 | dget(new_dentry); |
da1ce067 | 4288 | if (!is_dir || (flags & RENAME_EXCHANGE)) |
bc27027a MS |
4289 | lock_two_nondirectories(source, target); |
4290 | else if (target) | |
5955102c | 4291 | inode_lock(target); |
9055cba7 SW |
4292 | |
4293 | error = -EBUSY; | |
7af1364f | 4294 | if (is_local_mountpoint(old_dentry) || is_local_mountpoint(new_dentry)) |
9055cba7 SW |
4295 | goto out; |
4296 | ||
da1ce067 | 4297 | if (max_links && new_dir != old_dir) { |
bc27027a | 4298 | error = -EMLINK; |
da1ce067 | 4299 | if (is_dir && !new_is_dir && new_dir->i_nlink >= max_links) |
bc27027a | 4300 | goto out; |
da1ce067 MS |
4301 | if ((flags & RENAME_EXCHANGE) && !is_dir && new_is_dir && |
4302 | old_dir->i_nlink >= max_links) | |
4303 | goto out; | |
4304 | } | |
da1ce067 | 4305 | if (!is_dir) { |
bc27027a | 4306 | error = try_break_deleg(source, delegated_inode); |
8e6d782c BF |
4307 | if (error) |
4308 | goto out; | |
da1ce067 MS |
4309 | } |
4310 | if (target && !new_is_dir) { | |
4311 | error = try_break_deleg(target, delegated_inode); | |
4312 | if (error) | |
4313 | goto out; | |
8e6d782c | 4314 | } |
2773bf00 | 4315 | error = old_dir->i_op->rename(old_dir, old_dentry, |
18fc84da | 4316 | new_dir, new_dentry, flags); |
51892bbb SW |
4317 | if (error) |
4318 | goto out; | |
4319 | ||
da1ce067 | 4320 | if (!(flags & RENAME_EXCHANGE) && target) { |
8767712f AV |
4321 | if (is_dir) { |
4322 | shrink_dcache_parent(new_dentry); | |
bc27027a | 4323 | target->i_flags |= S_DEAD; |
8767712f | 4324 | } |
51892bbb | 4325 | dont_mount(new_dentry); |
8ed936b5 | 4326 | detach_mounts(new_dentry); |
bc27027a | 4327 | } |
da1ce067 MS |
4328 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) { |
4329 | if (!(flags & RENAME_EXCHANGE)) | |
4330 | d_move(old_dentry, new_dentry); | |
4331 | else | |
4332 | d_exchange(old_dentry, new_dentry); | |
4333 | } | |
51892bbb | 4334 | out: |
da1ce067 | 4335 | if (!is_dir || (flags & RENAME_EXCHANGE)) |
bc27027a MS |
4336 | unlock_two_nondirectories(source, target); |
4337 | else if (target) | |
5955102c | 4338 | inode_unlock(target); |
1da177e4 | 4339 | dput(new_dentry); |
da1ce067 | 4340 | if (!error) { |
f4ec3a3d | 4341 | fsnotify_move(old_dir, new_dir, &old_name.name, is_dir, |
da1ce067 MS |
4342 | !(flags & RENAME_EXCHANGE) ? target : NULL, old_dentry); |
4343 | if (flags & RENAME_EXCHANGE) { | |
f4ec3a3d | 4344 | fsnotify_move(new_dir, old_dir, &old_dentry->d_name, |
da1ce067 MS |
4345 | new_is_dir, NULL, new_dentry); |
4346 | } | |
4347 | } | |
49d31c2f | 4348 | release_dentry_name_snapshot(&old_name); |
0eeca283 | 4349 | |
1da177e4 LT |
4350 | return error; |
4351 | } | |
4d359507 | 4352 | EXPORT_SYMBOL(vfs_rename); |
1da177e4 | 4353 | |
e886663c JA |
4354 | int do_renameat2(int olddfd, struct filename *from, int newdfd, |
4355 | struct filename *to, unsigned int flags) | |
1da177e4 | 4356 | { |
2ad94ae6 AV |
4357 | struct dentry *old_dentry, *new_dentry; |
4358 | struct dentry *trap; | |
f5beed75 AV |
4359 | struct path old_path, new_path; |
4360 | struct qstr old_last, new_last; | |
4361 | int old_type, new_type; | |
8e6d782c | 4362 | struct inode *delegated_inode = NULL; |
f5beed75 | 4363 | unsigned int lookup_flags = 0, target_flags = LOOKUP_RENAME_TARGET; |
c6a94284 | 4364 | bool should_retry = false; |
e886663c | 4365 | int error = -EINVAL; |
520c8b16 | 4366 | |
0d7a8555 | 4367 | if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT)) |
e886663c | 4368 | goto put_both; |
da1ce067 | 4369 | |
0d7a8555 MS |
4370 | if ((flags & (RENAME_NOREPLACE | RENAME_WHITEOUT)) && |
4371 | (flags & RENAME_EXCHANGE)) | |
e886663c | 4372 | goto put_both; |
520c8b16 | 4373 | |
f5beed75 AV |
4374 | if (flags & RENAME_EXCHANGE) |
4375 | target_flags = 0; | |
4376 | ||
c6a94284 | 4377 | retry: |
e886663c JA |
4378 | from = filename_parentat(olddfd, from, lookup_flags, &old_path, |
4379 | &old_last, &old_type); | |
91a27b2a JL |
4380 | if (IS_ERR(from)) { |
4381 | error = PTR_ERR(from); | |
e886663c | 4382 | goto put_new; |
91a27b2a | 4383 | } |
1da177e4 | 4384 | |
e886663c JA |
4385 | to = filename_parentat(newdfd, to, lookup_flags, &new_path, &new_last, |
4386 | &new_type); | |
91a27b2a JL |
4387 | if (IS_ERR(to)) { |
4388 | error = PTR_ERR(to); | |
1da177e4 | 4389 | goto exit1; |
91a27b2a | 4390 | } |
1da177e4 LT |
4391 | |
4392 | error = -EXDEV; | |
f5beed75 | 4393 | if (old_path.mnt != new_path.mnt) |
1da177e4 LT |
4394 | goto exit2; |
4395 | ||
1da177e4 | 4396 | error = -EBUSY; |
f5beed75 | 4397 | if (old_type != LAST_NORM) |
1da177e4 LT |
4398 | goto exit2; |
4399 | ||
0a7c3937 MS |
4400 | if (flags & RENAME_NOREPLACE) |
4401 | error = -EEXIST; | |
f5beed75 | 4402 | if (new_type != LAST_NORM) |
1da177e4 LT |
4403 | goto exit2; |
4404 | ||
f5beed75 | 4405 | error = mnt_want_write(old_path.mnt); |
c30dabfe JK |
4406 | if (error) |
4407 | goto exit2; | |
4408 | ||
8e6d782c | 4409 | retry_deleg: |
f5beed75 | 4410 | trap = lock_rename(new_path.dentry, old_path.dentry); |
1da177e4 | 4411 | |
f5beed75 | 4412 | old_dentry = __lookup_hash(&old_last, old_path.dentry, lookup_flags); |
1da177e4 LT |
4413 | error = PTR_ERR(old_dentry); |
4414 | if (IS_ERR(old_dentry)) | |
4415 | goto exit3; | |
4416 | /* source must exist */ | |
4417 | error = -ENOENT; | |
b18825a7 | 4418 | if (d_is_negative(old_dentry)) |
1da177e4 | 4419 | goto exit4; |
f5beed75 | 4420 | new_dentry = __lookup_hash(&new_last, new_path.dentry, lookup_flags | target_flags); |
0a7c3937 MS |
4421 | error = PTR_ERR(new_dentry); |
4422 | if (IS_ERR(new_dentry)) | |
4423 | goto exit4; | |
4424 | error = -EEXIST; | |
4425 | if ((flags & RENAME_NOREPLACE) && d_is_positive(new_dentry)) | |
4426 | goto exit5; | |
da1ce067 MS |
4427 | if (flags & RENAME_EXCHANGE) { |
4428 | error = -ENOENT; | |
4429 | if (d_is_negative(new_dentry)) | |
4430 | goto exit5; | |
4431 | ||
4432 | if (!d_is_dir(new_dentry)) { | |
4433 | error = -ENOTDIR; | |
f5beed75 | 4434 | if (new_last.name[new_last.len]) |
da1ce067 MS |
4435 | goto exit5; |
4436 | } | |
4437 | } | |
1da177e4 | 4438 | /* unless the source is a directory trailing slashes give -ENOTDIR */ |
44b1d530 | 4439 | if (!d_is_dir(old_dentry)) { |
1da177e4 | 4440 | error = -ENOTDIR; |
f5beed75 | 4441 | if (old_last.name[old_last.len]) |
0a7c3937 | 4442 | goto exit5; |
f5beed75 | 4443 | if (!(flags & RENAME_EXCHANGE) && new_last.name[new_last.len]) |
0a7c3937 | 4444 | goto exit5; |
1da177e4 LT |
4445 | } |
4446 | /* source should not be ancestor of target */ | |
4447 | error = -EINVAL; | |
4448 | if (old_dentry == trap) | |
0a7c3937 | 4449 | goto exit5; |
1da177e4 | 4450 | /* target should not be an ancestor of source */ |
da1ce067 MS |
4451 | if (!(flags & RENAME_EXCHANGE)) |
4452 | error = -ENOTEMPTY; | |
1da177e4 LT |
4453 | if (new_dentry == trap) |
4454 | goto exit5; | |
4455 | ||
f5beed75 AV |
4456 | error = security_path_rename(&old_path, old_dentry, |
4457 | &new_path, new_dentry, flags); | |
be6d3e56 | 4458 | if (error) |
c30dabfe | 4459 | goto exit5; |
f5beed75 AV |
4460 | error = vfs_rename(old_path.dentry->d_inode, old_dentry, |
4461 | new_path.dentry->d_inode, new_dentry, | |
520c8b16 | 4462 | &delegated_inode, flags); |
1da177e4 LT |
4463 | exit5: |
4464 | dput(new_dentry); | |
4465 | exit4: | |
4466 | dput(old_dentry); | |
4467 | exit3: | |
f5beed75 | 4468 | unlock_rename(new_path.dentry, old_path.dentry); |
8e6d782c BF |
4469 | if (delegated_inode) { |
4470 | error = break_deleg_wait(&delegated_inode); | |
4471 | if (!error) | |
4472 | goto retry_deleg; | |
4473 | } | |
f5beed75 | 4474 | mnt_drop_write(old_path.mnt); |
1da177e4 | 4475 | exit2: |
c6a94284 JL |
4476 | if (retry_estale(error, lookup_flags)) |
4477 | should_retry = true; | |
f5beed75 | 4478 | path_put(&new_path); |
1da177e4 | 4479 | exit1: |
f5beed75 | 4480 | path_put(&old_path); |
c6a94284 JL |
4481 | if (should_retry) { |
4482 | should_retry = false; | |
4483 | lookup_flags |= LOOKUP_REVAL; | |
4484 | goto retry; | |
4485 | } | |
e886663c JA |
4486 | put_both: |
4487 | if (!IS_ERR(from)) | |
4488 | putname(from); | |
4489 | put_new: | |
4490 | if (!IS_ERR(to)) | |
4491 | putname(to); | |
1da177e4 LT |
4492 | return error; |
4493 | } | |
4494 | ||
ee81feb6 DB |
4495 | SYSCALL_DEFINE5(renameat2, int, olddfd, const char __user *, oldname, |
4496 | int, newdfd, const char __user *, newname, unsigned int, flags) | |
4497 | { | |
e886663c JA |
4498 | return do_renameat2(olddfd, getname(oldname), newdfd, getname(newname), |
4499 | flags); | |
ee81feb6 DB |
4500 | } |
4501 | ||
520c8b16 MS |
4502 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
4503 | int, newdfd, const char __user *, newname) | |
4504 | { | |
e886663c JA |
4505 | return do_renameat2(olddfd, getname(oldname), newdfd, getname(newname), |
4506 | 0); | |
520c8b16 MS |
4507 | } |
4508 | ||
a26eab24 | 4509 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4510 | { |
e886663c JA |
4511 | return do_renameat2(AT_FDCWD, getname(oldname), AT_FDCWD, |
4512 | getname(newname), 0); | |
5590ff0d UD |
4513 | } |
4514 | ||
5d826c84 | 4515 | int readlink_copy(char __user *buffer, int buflen, const char *link) |
1da177e4 | 4516 | { |
5d826c84 | 4517 | int len = PTR_ERR(link); |
1da177e4 LT |
4518 | if (IS_ERR(link)) |
4519 | goto out; | |
4520 | ||
4521 | len = strlen(link); | |
4522 | if (len > (unsigned) buflen) | |
4523 | len = buflen; | |
4524 | if (copy_to_user(buffer, link, len)) | |
4525 | len = -EFAULT; | |
4526 | out: | |
4527 | return len; | |
4528 | } | |
4529 | ||
fd4a0edf MS |
4530 | /** |
4531 | * vfs_readlink - copy symlink body into userspace buffer | |
4532 | * @dentry: dentry on which to get symbolic link | |
4533 | * @buffer: user memory pointer | |
4534 | * @buflen: size of buffer | |
4535 | * | |
4536 | * Does not touch atime. That's up to the caller if necessary | |
4537 | * | |
4538 | * Does not call security hook. | |
4539 | */ | |
4540 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
4541 | { | |
4542 | struct inode *inode = d_inode(dentry); | |
f2df5da6 AV |
4543 | DEFINE_DELAYED_CALL(done); |
4544 | const char *link; | |
4545 | int res; | |
fd4a0edf | 4546 | |
76fca90e MS |
4547 | if (unlikely(!(inode->i_opflags & IOP_DEFAULT_READLINK))) { |
4548 | if (unlikely(inode->i_op->readlink)) | |
4549 | return inode->i_op->readlink(dentry, buffer, buflen); | |
4550 | ||
4551 | if (!d_is_symlink(dentry)) | |
4552 | return -EINVAL; | |
4553 | ||
4554 | spin_lock(&inode->i_lock); | |
4555 | inode->i_opflags |= IOP_DEFAULT_READLINK; | |
4556 | spin_unlock(&inode->i_lock); | |
4557 | } | |
fd4a0edf | 4558 | |
4c4f7c19 | 4559 | link = READ_ONCE(inode->i_link); |
f2df5da6 AV |
4560 | if (!link) { |
4561 | link = inode->i_op->get_link(dentry, inode, &done); | |
4562 | if (IS_ERR(link)) | |
4563 | return PTR_ERR(link); | |
4564 | } | |
4565 | res = readlink_copy(buffer, buflen, link); | |
4566 | do_delayed_call(&done); | |
4567 | return res; | |
fd4a0edf MS |
4568 | } |
4569 | EXPORT_SYMBOL(vfs_readlink); | |
1da177e4 | 4570 | |
d60874cd MS |
4571 | /** |
4572 | * vfs_get_link - get symlink body | |
4573 | * @dentry: dentry on which to get symbolic link | |
4574 | * @done: caller needs to free returned data with this | |
4575 | * | |
4576 | * Calls security hook and i_op->get_link() on the supplied inode. | |
4577 | * | |
4578 | * It does not touch atime. That's up to the caller if necessary. | |
4579 | * | |
4580 | * Does not work on "special" symlinks like /proc/$$/fd/N | |
4581 | */ | |
4582 | const char *vfs_get_link(struct dentry *dentry, struct delayed_call *done) | |
4583 | { | |
4584 | const char *res = ERR_PTR(-EINVAL); | |
4585 | struct inode *inode = d_inode(dentry); | |
4586 | ||
4587 | if (d_is_symlink(dentry)) { | |
4588 | res = ERR_PTR(security_inode_readlink(dentry)); | |
4589 | if (!res) | |
4590 | res = inode->i_op->get_link(dentry, inode, done); | |
4591 | } | |
4592 | return res; | |
4593 | } | |
4594 | EXPORT_SYMBOL(vfs_get_link); | |
4595 | ||
1da177e4 | 4596 | /* get the link contents into pagecache */ |
6b255391 | 4597 | const char *page_get_link(struct dentry *dentry, struct inode *inode, |
fceef393 | 4598 | struct delayed_call *callback) |
1da177e4 | 4599 | { |
ebd09abb DG |
4600 | char *kaddr; |
4601 | struct page *page; | |
6b255391 AV |
4602 | struct address_space *mapping = inode->i_mapping; |
4603 | ||
d3883d4f AV |
4604 | if (!dentry) { |
4605 | page = find_get_page(mapping, 0); | |
4606 | if (!page) | |
4607 | return ERR_PTR(-ECHILD); | |
4608 | if (!PageUptodate(page)) { | |
4609 | put_page(page); | |
4610 | return ERR_PTR(-ECHILD); | |
4611 | } | |
4612 | } else { | |
4613 | page = read_mapping_page(mapping, 0, NULL); | |
4614 | if (IS_ERR(page)) | |
4615 | return (char*)page; | |
4616 | } | |
fceef393 | 4617 | set_delayed_call(callback, page_put_link, page); |
21fc61c7 AV |
4618 | BUG_ON(mapping_gfp_mask(mapping) & __GFP_HIGHMEM); |
4619 | kaddr = page_address(page); | |
6b255391 | 4620 | nd_terminate_link(kaddr, inode->i_size, PAGE_SIZE - 1); |
ebd09abb | 4621 | return kaddr; |
1da177e4 LT |
4622 | } |
4623 | ||
6b255391 | 4624 | EXPORT_SYMBOL(page_get_link); |
1da177e4 | 4625 | |
fceef393 | 4626 | void page_put_link(void *arg) |
1da177e4 | 4627 | { |
fceef393 | 4628 | put_page(arg); |
1da177e4 | 4629 | } |
4d359507 | 4630 | EXPORT_SYMBOL(page_put_link); |
1da177e4 | 4631 | |
aa80deab AV |
4632 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) |
4633 | { | |
fceef393 | 4634 | DEFINE_DELAYED_CALL(done); |
6b255391 AV |
4635 | int res = readlink_copy(buffer, buflen, |
4636 | page_get_link(dentry, d_inode(dentry), | |
fceef393 AV |
4637 | &done)); |
4638 | do_delayed_call(&done); | |
aa80deab AV |
4639 | return res; |
4640 | } | |
4641 | EXPORT_SYMBOL(page_readlink); | |
4642 | ||
54566b2c NP |
4643 | /* |
4644 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
4645 | */ | |
4646 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
4647 | { |
4648 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 4649 | struct page *page; |
afddba49 | 4650 | void *fsdata; |
beb497ab | 4651 | int err; |
c718a975 | 4652 | unsigned int flags = 0; |
54566b2c NP |
4653 | if (nofs) |
4654 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 4655 | |
7e53cac4 | 4656 | retry: |
afddba49 | 4657 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 4658 | flags, &page, &fsdata); |
1da177e4 | 4659 | if (err) |
afddba49 NP |
4660 | goto fail; |
4661 | ||
21fc61c7 | 4662 | memcpy(page_address(page), symname, len-1); |
afddba49 NP |
4663 | |
4664 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
4665 | page, fsdata); | |
1da177e4 LT |
4666 | if (err < 0) |
4667 | goto fail; | |
afddba49 NP |
4668 | if (err < len-1) |
4669 | goto retry; | |
4670 | ||
1da177e4 LT |
4671 | mark_inode_dirty(inode); |
4672 | return 0; | |
1da177e4 LT |
4673 | fail: |
4674 | return err; | |
4675 | } | |
4d359507 | 4676 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 | 4677 | |
0adb25d2 KK |
4678 | int page_symlink(struct inode *inode, const char *symname, int len) |
4679 | { | |
4680 | return __page_symlink(inode, symname, len, | |
c62d2555 | 4681 | !mapping_gfp_constraint(inode->i_mapping, __GFP_FS)); |
0adb25d2 | 4682 | } |
4d359507 | 4683 | EXPORT_SYMBOL(page_symlink); |
0adb25d2 | 4684 | |
92e1d5be | 4685 | const struct inode_operations page_symlink_inode_operations = { |
6b255391 | 4686 | .get_link = page_get_link, |
1da177e4 | 4687 | }; |
1da177e4 | 4688 | EXPORT_SYMBOL(page_symlink_inode_operations); |