1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright (C) 1991, 1992 Linus Torvalds
6 * changes by Thomas Schoebel-Theuer
9 #include <linux/export.h>
10 #include <linux/time.h>
12 #include <linux/string.h>
13 #include <linux/sched/signal.h>
14 #include <linux/capability.h>
15 #include <linux/fsnotify.h>
16 #include <linux/fcntl.h>
17 #include <linux/security.h>
18 #include <linux/evm.h>
19 #include <linux/ima.h>
24 * setattr_should_drop_sgid - determine whether the setgid bit needs to be
26 * @idmap: idmap of the mount @inode was found from
27 * @inode: inode to check
29 * This function determines whether the setgid bit needs to be removed.
30 * We retain backwards compatibility and require setgid bit to be removed
31 * unconditionally if S_IXGRP is set. Otherwise we have the exact same
32 * requirements as setattr_prepare() and setattr_copy().
34 * Return: ATTR_KILL_SGID if setgid bit needs to be removed, 0 otherwise.
36 int setattr_should_drop_sgid(struct mnt_idmap *idmap,
37 const struct inode *inode)
39 struct user_namespace *mnt_userns = mnt_idmap_owner(idmap);
40 umode_t mode = inode->i_mode;
42 if (!(mode & S_ISGID))
45 return ATTR_KILL_SGID;
46 if (!in_group_or_capable(idmap, inode,
47 i_gid_into_vfsgid(mnt_userns, inode)))
48 return ATTR_KILL_SGID;
53 * setattr_should_drop_suidgid - determine whether the set{g,u}id bit needs to
55 * @idmap: idmap of the mount @inode was found from
56 * @inode: inode to check
58 * This function determines whether the set{g,u}id bits need to be removed.
59 * If the setuid bit needs to be removed ATTR_KILL_SUID is returned. If the
60 * setgid bit needs to be removed ATTR_KILL_SGID is returned. If both
61 * set{g,u}id bits need to be removed the corresponding mask of both flags is
64 * Return: A mask of ATTR_KILL_S{G,U}ID indicating which - if any - setid bits
65 * to remove, 0 otherwise.
67 int setattr_should_drop_suidgid(struct mnt_idmap *idmap,
70 umode_t mode = inode->i_mode;
73 /* suid always must be killed */
74 if (unlikely(mode & S_ISUID))
75 kill = ATTR_KILL_SUID;
77 kill |= setattr_should_drop_sgid(idmap, inode);
79 if (unlikely(kill && !capable(CAP_FSETID) && S_ISREG(mode)))
84 EXPORT_SYMBOL(setattr_should_drop_suidgid);
87 * chown_ok - verify permissions to chown inode
88 * @idmap: idmap of the mount @inode was found from
89 * @inode: inode to check permissions on
90 * @ia_vfsuid: uid to chown @inode to
92 * If the inode has been found through an idmapped mount the idmap of
93 * the vfsmount must be passed through @idmap. This function will then
94 * take care to map the inode according to @idmap before checking
95 * permissions. On non-idmapped mounts or if permission checking is to be
96 * performed on the raw inode simply pass @nop_mnt_idmap.
98 static bool chown_ok(struct mnt_idmap *idmap,
99 const struct inode *inode, vfsuid_t ia_vfsuid)
101 struct user_namespace *mnt_userns = mnt_idmap_owner(idmap);
103 vfsuid_t vfsuid = i_uid_into_vfsuid(mnt_userns, inode);
104 if (vfsuid_eq_kuid(vfsuid, current_fsuid()) &&
105 vfsuid_eq(ia_vfsuid, vfsuid))
107 if (capable_wrt_inode_uidgid(idmap, inode, CAP_CHOWN))
109 if (!vfsuid_valid(vfsuid) &&
110 ns_capable(inode->i_sb->s_user_ns, CAP_CHOWN))
116 * chgrp_ok - verify permissions to chgrp inode
117 * @idmap: idmap of the mount @inode was found from
118 * @inode: inode to check permissions on
119 * @ia_vfsgid: gid to chown @inode to
121 * If the inode has been found through an idmapped mount the idmap of
122 * the vfsmount must be passed through @idmap. This function will then
123 * take care to map the inode according to @idmap before checking
124 * permissions. On non-idmapped mounts or if permission checking is to be
125 * performed on the raw inode simply pass @nop_mnt_idmap.
127 static bool chgrp_ok(struct mnt_idmap *idmap,
128 const struct inode *inode, vfsgid_t ia_vfsgid)
130 struct user_namespace *mnt_userns = mnt_idmap_owner(idmap);
132 vfsgid_t vfsgid = i_gid_into_vfsgid(mnt_userns, inode);
133 vfsuid_t vfsuid = i_uid_into_vfsuid(mnt_userns, inode);
134 if (vfsuid_eq_kuid(vfsuid, current_fsuid())) {
135 if (vfsgid_eq(ia_vfsgid, vfsgid))
137 if (vfsgid_in_group_p(ia_vfsgid))
140 if (capable_wrt_inode_uidgid(idmap, inode, CAP_CHOWN))
142 if (!vfsgid_valid(vfsgid) &&
143 ns_capable(inode->i_sb->s_user_ns, CAP_CHOWN))
149 * setattr_prepare - check if attribute changes to a dentry are allowed
150 * @idmap: idmap of the mount the inode was found from
151 * @dentry: dentry to check
152 * @attr: attributes to change
154 * Check if we are allowed to change the attributes contained in @attr
155 * in the given dentry. This includes the normal unix access permission
156 * checks, as well as checks for rlimits and others. The function also clears
157 * SGID bit from mode if user is not allowed to set it. Also file capabilities
158 * and IMA extended attributes are cleared if ATTR_KILL_PRIV is set.
160 * If the inode has been found through an idmapped mount the idmap of
161 * the vfsmount must be passed through @idmap. This function will then
162 * take care to map the inode according to @idmap before checking
163 * permissions. On non-idmapped mounts or if permission checking is to be
164 * performed on the raw inode simply passs @nop_mnt_idmap.
166 * Should be called as the first thing in ->setattr implementations,
167 * possibly after taking additional locks.
169 int setattr_prepare(struct mnt_idmap *idmap, struct dentry *dentry,
172 struct user_namespace *mnt_userns = mnt_idmap_owner(idmap);
173 struct inode *inode = d_inode(dentry);
174 unsigned int ia_valid = attr->ia_valid;
177 * First check size constraints. These can't be overriden using
180 if (ia_valid & ATTR_SIZE) {
181 int error = inode_newsize_ok(inode, attr->ia_size);
186 /* If force is set do it anyway. */
187 if (ia_valid & ATTR_FORCE)
190 /* Make sure a caller can chown. */
191 if ((ia_valid & ATTR_UID) &&
192 !chown_ok(idmap, inode, attr->ia_vfsuid))
195 /* Make sure caller can chgrp. */
196 if ((ia_valid & ATTR_GID) &&
197 !chgrp_ok(idmap, inode, attr->ia_vfsgid))
200 /* Make sure a caller can chmod. */
201 if (ia_valid & ATTR_MODE) {
204 if (!inode_owner_or_capable(idmap, inode))
207 if (ia_valid & ATTR_GID)
208 vfsgid = attr->ia_vfsgid;
210 vfsgid = i_gid_into_vfsgid(mnt_userns, inode);
212 /* Also check the setgid bit! */
213 if (!in_group_or_capable(idmap, inode, vfsgid))
214 attr->ia_mode &= ~S_ISGID;
217 /* Check for setting the inode time. */
218 if (ia_valid & (ATTR_MTIME_SET | ATTR_ATIME_SET | ATTR_TIMES_SET)) {
219 if (!inode_owner_or_capable(idmap, inode))
224 /* User has permission for the change */
225 if (ia_valid & ATTR_KILL_PRIV) {
228 error = security_inode_killpriv(idmap, dentry);
235 EXPORT_SYMBOL(setattr_prepare);
238 * inode_newsize_ok - may this inode be truncated to a given size
239 * @inode: the inode to be truncated
240 * @offset: the new size to assign to the inode
242 * inode_newsize_ok must be called with i_mutex held.
244 * inode_newsize_ok will check filesystem limits and ulimits to check that the
245 * new inode size is within limits. inode_newsize_ok will also send SIGXFSZ
246 * when necessary. Caller must not proceed with inode size change if failure is
247 * returned. @inode must be a file (not directory), with appropriate
248 * permissions to allow truncate (inode_newsize_ok does NOT check these
251 * Return: 0 on success, -ve errno on failure
253 int inode_newsize_ok(const struct inode *inode, loff_t offset)
257 if (inode->i_size < offset) {
260 limit = rlimit(RLIMIT_FSIZE);
261 if (limit != RLIM_INFINITY && offset > limit)
263 if (offset > inode->i_sb->s_maxbytes)
267 * truncation of in-use swapfiles is disallowed - it would
268 * cause subsequent swapout to scribble on the now-freed
271 if (IS_SWAPFILE(inode))
277 send_sig(SIGXFSZ, current, 0);
281 EXPORT_SYMBOL(inode_newsize_ok);
284 * setattr_copy - copy simple metadata updates into the generic inode
285 * @idmap: idmap of the mount the inode was found from
286 * @inode: the inode to be updated
287 * @attr: the new attributes
289 * setattr_copy must be called with i_mutex held.
291 * setattr_copy updates the inode's metadata with that specified
292 * in attr on idmapped mounts. Necessary permission checks to determine
293 * whether or not the S_ISGID property needs to be removed are performed with
294 * the correct idmapped mount permission helpers.
295 * Noticeably missing is inode size update, which is more complex
296 * as it requires pagecache updates.
298 * If the inode has been found through an idmapped mount the idmap of
299 * the vfsmount must be passed through @idmap. This function will then
300 * take care to map the inode according to @idmap before checking
301 * permissions. On non-idmapped mounts or if permission checking is to be
302 * performed on the raw inode simply pass @nop_mnt_idmap.
304 * The inode is not marked as dirty after this operation. The rationale is
305 * that for "simple" filesystems, the struct inode is the inode storage.
306 * The caller is free to mark the inode dirty afterwards if needed.
308 void setattr_copy(struct mnt_idmap *idmap, struct inode *inode,
309 const struct iattr *attr)
311 struct user_namespace *mnt_userns = mnt_idmap_owner(idmap);
312 unsigned int ia_valid = attr->ia_valid;
314 i_uid_update(mnt_userns, attr, inode);
315 i_gid_update(mnt_userns, attr, inode);
316 if (ia_valid & ATTR_ATIME)
317 inode->i_atime = attr->ia_atime;
318 if (ia_valid & ATTR_MTIME)
319 inode->i_mtime = attr->ia_mtime;
320 if (ia_valid & ATTR_CTIME)
321 inode->i_ctime = attr->ia_ctime;
322 if (ia_valid & ATTR_MODE) {
323 umode_t mode = attr->ia_mode;
324 if (!in_group_or_capable(idmap, inode,
325 i_gid_into_vfsgid(mnt_userns, inode)))
327 inode->i_mode = mode;
330 EXPORT_SYMBOL(setattr_copy);
332 int may_setattr(struct mnt_idmap *idmap, struct inode *inode,
333 unsigned int ia_valid)
337 if (ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID | ATTR_TIMES_SET)) {
338 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
343 * If utimes(2) and friends are called with times == NULL (or both
344 * times are UTIME_NOW), then we need to check for write permission
346 if (ia_valid & ATTR_TOUCH) {
347 if (IS_IMMUTABLE(inode))
350 if (!inode_owner_or_capable(idmap, inode)) {
351 error = inode_permission(idmap, inode, MAY_WRITE);
358 EXPORT_SYMBOL(may_setattr);
361 * notify_change - modify attributes of a filesytem object
362 * @idmap: idmap of the mount the inode was found from
363 * @dentry: object affected
364 * @attr: new attributes
365 * @delegated_inode: returns inode, if the inode is delegated
367 * The caller must hold the i_mutex on the affected object.
369 * If notify_change discovers a delegation in need of breaking,
370 * it will return -EWOULDBLOCK and return a reference to the inode in
371 * delegated_inode. The caller should then break the delegation and
372 * retry. Because breaking a delegation may take a long time, the
373 * caller should drop the i_mutex before doing so.
375 * Alternatively, a caller may pass NULL for delegated_inode. This may
376 * be appropriate for callers that expect the underlying filesystem not
377 * to be NFS exported. Also, passing NULL is fine for callers holding
378 * the file open for write, as there can be no conflicting delegation in
381 * If the inode has been found through an idmapped mount the idmap of
382 * the vfsmount must be passed through @idmap. This function will then
383 * take care to map the inode according to @idmap before checking
384 * permissions. On non-idmapped mounts or if permission checking is to be
385 * performed on the raw inode simply pass @nop_mnt_idmap.
387 int notify_change(struct mnt_idmap *idmap, struct dentry *dentry,
388 struct iattr *attr, struct inode **delegated_inode)
390 struct user_namespace *mnt_userns = mnt_idmap_owner(idmap);
391 struct inode *inode = dentry->d_inode;
392 umode_t mode = inode->i_mode;
394 struct timespec64 now;
395 unsigned int ia_valid = attr->ia_valid;
397 WARN_ON_ONCE(!inode_is_locked(inode));
399 error = may_setattr(idmap, inode, ia_valid);
403 if ((ia_valid & ATTR_MODE)) {
404 umode_t amode = attr->ia_mode;
405 /* Flag setting protected by i_mutex */
407 inode->i_flags &= ~S_NOSEC;
410 now = current_time(inode);
412 attr->ia_ctime = now;
413 if (!(ia_valid & ATTR_ATIME_SET))
414 attr->ia_atime = now;
416 attr->ia_atime = timestamp_truncate(attr->ia_atime, inode);
417 if (!(ia_valid & ATTR_MTIME_SET))
418 attr->ia_mtime = now;
420 attr->ia_mtime = timestamp_truncate(attr->ia_mtime, inode);
422 if (ia_valid & ATTR_KILL_PRIV) {
423 error = security_inode_need_killpriv(dentry);
427 ia_valid = attr->ia_valid &= ~ATTR_KILL_PRIV;
431 * We now pass ATTR_KILL_S*ID to the lower level setattr function so
432 * that the function has the ability to reinterpret a mode change
433 * that's due to these bits. This adds an implicit restriction that
434 * no function will ever call notify_change with both ATTR_MODE and
435 * ATTR_KILL_S*ID set.
437 if ((ia_valid & (ATTR_KILL_SUID|ATTR_KILL_SGID)) &&
438 (ia_valid & ATTR_MODE))
441 if (ia_valid & ATTR_KILL_SUID) {
442 if (mode & S_ISUID) {
443 ia_valid = attr->ia_valid |= ATTR_MODE;
444 attr->ia_mode = (inode->i_mode & ~S_ISUID);
447 if (ia_valid & ATTR_KILL_SGID) {
448 if (mode & S_ISGID) {
449 if (!(ia_valid & ATTR_MODE)) {
450 ia_valid = attr->ia_valid |= ATTR_MODE;
451 attr->ia_mode = inode->i_mode;
453 attr->ia_mode &= ~S_ISGID;
456 if (!(attr->ia_valid & ~(ATTR_KILL_SUID | ATTR_KILL_SGID)))
460 * Verify that uid/gid changes are valid in the target
461 * namespace of the superblock.
463 if (ia_valid & ATTR_UID &&
464 !vfsuid_has_fsmapping(mnt_userns, inode->i_sb->s_user_ns,
467 if (ia_valid & ATTR_GID &&
468 !vfsgid_has_fsmapping(mnt_userns, inode->i_sb->s_user_ns,
472 /* Don't allow modifications of files with invalid uids or
473 * gids unless those uids & gids are being made valid.
475 if (!(ia_valid & ATTR_UID) &&
476 !vfsuid_valid(i_uid_into_vfsuid(mnt_userns, inode)))
478 if (!(ia_valid & ATTR_GID) &&
479 !vfsgid_valid(i_gid_into_vfsgid(mnt_userns, inode)))
482 error = security_inode_setattr(idmap, dentry, attr);
485 error = try_break_deleg(inode, delegated_inode);
489 if (inode->i_op->setattr)
490 error = inode->i_op->setattr(idmap, dentry, attr);
492 error = simple_setattr(idmap, dentry, attr);
495 fsnotify_change(dentry, ia_valid);
496 ima_inode_post_setattr(idmap, dentry);
497 evm_inode_post_setattr(dentry, ia_valid);
502 EXPORT_SYMBOL(notify_change);