4 Extended attribute handling.
6 Copyright (C) 2001 by Andreas Gruenbacher <a.gruenbacher@computer.org>
7 Copyright (C) 2001 SGI - Silicon Graphics, Inc <linux-xfs@oss.sgi.com>
8 Copyright (c) 2004 Red Hat, Inc., James Morris <jmorris@redhat.com>
11 #include <linux/slab.h>
12 #include <linux/file.h>
13 #include <linux/xattr.h>
14 #include <linux/mount.h>
15 #include <linux/namei.h>
16 #include <linux/security.h>
17 #include <linux/evm.h>
18 #include <linux/syscalls.h>
19 #include <linux/export.h>
20 #include <linux/fsnotify.h>
21 #include <linux/audit.h>
22 #include <linux/vmalloc.h>
23 #include <linux/posix_acl_xattr.h>
25 #include <linux/uaccess.h>
28 strcmp_prefix(const char *a, const char *a_prefix)
30 while (*a_prefix && *a == *a_prefix) {
34 return *a_prefix ? NULL : a;
38 * In order to implement different sets of xattr operations for each xattr
39 * prefix, a filesystem should create a null-terminated array of struct
40 * xattr_handler (one for each prefix) and hang a pointer to it off of the
41 * s_xattr field of the superblock.
43 #define for_each_xattr_handler(handlers, handler) \
45 for ((handler) = *(handlers)++; \
47 (handler) = *(handlers)++)
50 * Find the xattr_handler with the matching prefix.
52 static const struct xattr_handler *
53 xattr_resolve_name(struct inode *inode, const char **name)
55 const struct xattr_handler **handlers = inode->i_sb->s_xattr;
56 const struct xattr_handler *handler;
58 if (!(inode->i_opflags & IOP_XATTR)) {
59 if (unlikely(is_bad_inode(inode)))
61 return ERR_PTR(-EOPNOTSUPP);
63 for_each_xattr_handler(handlers, handler) {
66 n = strcmp_prefix(*name, xattr_prefix(handler));
68 if (!handler->prefix ^ !*n) {
71 return ERR_PTR(-EINVAL);
77 return ERR_PTR(-EOPNOTSUPP);
81 * Check permissions for extended attribute access. This is a bit complicated
82 * because different namespaces have very different rules.
85 xattr_permission(struct inode *inode, const char *name, int mask)
88 * We can never set or remove an extended attribute on a read-only
89 * filesystem or on an immutable / append-only inode.
91 if (mask & MAY_WRITE) {
92 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
95 * Updating an xattr will likely cause i_uid and i_gid
96 * to be writen back improperly if their true value is
99 if (HAS_UNMAPPED_ID(inode))
104 * No restriction for security.* and system.* from the VFS. Decision
105 * on these is left to the underlying filesystem / security module.
107 if (!strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN) ||
108 !strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
112 * The trusted.* namespace can only be accessed by privileged users.
114 if (!strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN)) {
115 if (!capable(CAP_SYS_ADMIN))
116 return (mask & MAY_WRITE) ? -EPERM : -ENODATA;
121 * In the user.* namespace, only regular files and directories can have
122 * extended attributes. For sticky directories, only the owner and
123 * privileged users can write attributes.
125 if (!strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN)) {
126 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
127 return (mask & MAY_WRITE) ? -EPERM : -ENODATA;
128 if (S_ISDIR(inode->i_mode) && (inode->i_mode & S_ISVTX) &&
129 (mask & MAY_WRITE) && !inode_owner_or_capable(inode))
133 return inode_permission(inode, mask);
137 __vfs_setxattr(struct dentry *dentry, struct inode *inode, const char *name,
138 const void *value, size_t size, int flags)
140 const struct xattr_handler *handler;
142 handler = xattr_resolve_name(inode, &name);
144 return PTR_ERR(handler);
148 value = ""; /* empty EA, do not remove */
149 return handler->set(handler, dentry, inode, name, value, size, flags);
151 EXPORT_SYMBOL(__vfs_setxattr);
154 * __vfs_setxattr_noperm - perform setxattr operation without performing
157 * @dentry - object to perform setxattr on
158 * @name - xattr name to set
159 * @value - value to set @name to
160 * @size - size of @value
161 * @flags - flags to pass into filesystem operations
163 * returns the result of the internal setxattr or setsecurity operations.
165 * This function requires the caller to lock the inode's i_mutex before it
166 * is executed. It also assumes that the caller will make the appropriate
169 int __vfs_setxattr_noperm(struct dentry *dentry, const char *name,
170 const void *value, size_t size, int flags)
172 struct inode *inode = dentry->d_inode;
174 int issec = !strncmp(name, XATTR_SECURITY_PREFIX,
175 XATTR_SECURITY_PREFIX_LEN);
178 inode->i_flags &= ~S_NOSEC;
179 if (inode->i_opflags & IOP_XATTR) {
180 error = __vfs_setxattr(dentry, inode, name, value, size, flags);
182 fsnotify_xattr(dentry);
183 security_inode_post_setxattr(dentry, name, value,
187 if (unlikely(is_bad_inode(inode)))
190 if (error == -EAGAIN) {
194 const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
196 error = security_inode_setsecurity(inode, suffix, value,
199 fsnotify_xattr(dentry);
207 * __vfs_setxattr_locked: set an extended attribute while holding the inode
210 * @dentry - object to perform setxattr on
211 * @name - xattr name to set
212 * @value - value to set @name to
213 * @size - size of @value
214 * @flags - flags to pass into filesystem operations
215 * @delegated_inode - on return, will contain an inode pointer that
216 * a delegation was broken on, NULL if none.
219 __vfs_setxattr_locked(struct dentry *dentry, const char *name,
220 const void *value, size_t size, int flags,
221 struct inode **delegated_inode)
223 struct inode *inode = dentry->d_inode;
226 error = xattr_permission(inode, name, MAY_WRITE);
230 error = security_inode_setxattr(dentry, name, value, size, flags);
234 error = try_break_deleg(inode, delegated_inode);
238 error = __vfs_setxattr_noperm(dentry, name, value, size, flags);
243 EXPORT_SYMBOL_GPL(__vfs_setxattr_locked);
246 vfs_setxattr(struct dentry *dentry, const char *name, const void *value,
247 size_t size, int flags)
249 struct inode *inode = dentry->d_inode;
250 struct inode *delegated_inode = NULL;
255 error = __vfs_setxattr_locked(dentry, name, value, size, flags,
259 if (delegated_inode) {
260 error = break_deleg_wait(&delegated_inode);
266 EXPORT_SYMBOL_GPL(vfs_setxattr);
269 xattr_getsecurity(struct inode *inode, const char *name, void *value,
275 if (!value || !size) {
276 len = security_inode_getsecurity(inode, name, &buffer, false);
280 len = security_inode_getsecurity(inode, name, &buffer, true);
287 memcpy(value, buffer, len);
295 * vfs_getxattr_alloc - allocate memory, if necessary, before calling getxattr
297 * Allocate memory, if not already allocated, or re-allocate correct size,
298 * before retrieving the extended attribute.
300 * Returns the result of alloc, if failed, or the getxattr operation.
303 vfs_getxattr_alloc(struct dentry *dentry, const char *name, char **xattr_value,
304 size_t xattr_size, gfp_t flags)
306 const struct xattr_handler *handler;
307 struct inode *inode = dentry->d_inode;
308 char *value = *xattr_value;
311 error = xattr_permission(inode, name, MAY_READ);
315 handler = xattr_resolve_name(inode, &name);
317 return PTR_ERR(handler);
320 error = handler->get(handler, dentry, inode, name, NULL, 0);
324 if (!value || (error > xattr_size)) {
325 value = krealloc(*xattr_value, error + 1, flags);
328 memset(value, 0, error + 1);
331 error = handler->get(handler, dentry, inode, name, value, error);
332 *xattr_value = value;
337 __vfs_getxattr(struct dentry *dentry, struct inode *inode, const char *name,
338 void *value, size_t size)
340 const struct xattr_handler *handler;
342 handler = xattr_resolve_name(inode, &name);
344 return PTR_ERR(handler);
347 return handler->get(handler, dentry, inode, name, value, size);
349 EXPORT_SYMBOL(__vfs_getxattr);
352 vfs_getxattr(struct dentry *dentry, const char *name, void *value, size_t size)
354 struct inode *inode = dentry->d_inode;
357 error = xattr_permission(inode, name, MAY_READ);
361 error = security_inode_getxattr(dentry, name);
365 if (!strncmp(name, XATTR_SECURITY_PREFIX,
366 XATTR_SECURITY_PREFIX_LEN)) {
367 const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
368 int ret = xattr_getsecurity(inode, suffix, value, size);
370 * Only overwrite the return value if a security module
371 * is actually active.
373 if (ret == -EOPNOTSUPP)
378 return __vfs_getxattr(dentry, inode, name, value, size);
380 EXPORT_SYMBOL_GPL(vfs_getxattr);
383 vfs_listxattr(struct dentry *dentry, char *list, size_t size)
385 struct inode *inode = d_inode(dentry);
388 error = security_inode_listxattr(dentry);
391 if (inode->i_op->listxattr && (inode->i_opflags & IOP_XATTR)) {
392 error = inode->i_op->listxattr(dentry, list, size);
394 error = security_inode_listsecurity(inode, list, size);
395 if (size && error > size)
400 EXPORT_SYMBOL_GPL(vfs_listxattr);
403 __vfs_removexattr(struct dentry *dentry, const char *name)
405 struct inode *inode = d_inode(dentry);
406 const struct xattr_handler *handler;
408 handler = xattr_resolve_name(inode, &name);
410 return PTR_ERR(handler);
413 return handler->set(handler, dentry, inode, name, NULL, 0, XATTR_REPLACE);
415 EXPORT_SYMBOL(__vfs_removexattr);
418 * __vfs_removexattr_locked: set an extended attribute while holding the inode
421 * @dentry - object to perform setxattr on
422 * @name - name of xattr to remove
423 * @delegated_inode - on return, will contain an inode pointer that
424 * a delegation was broken on, NULL if none.
427 __vfs_removexattr_locked(struct dentry *dentry, const char *name,
428 struct inode **delegated_inode)
430 struct inode *inode = dentry->d_inode;
433 error = xattr_permission(inode, name, MAY_WRITE);
437 error = security_inode_removexattr(dentry, name);
441 error = try_break_deleg(inode, delegated_inode);
445 error = __vfs_removexattr(dentry, name);
448 fsnotify_xattr(dentry);
449 evm_inode_post_removexattr(dentry, name);
455 EXPORT_SYMBOL_GPL(__vfs_removexattr_locked);
458 vfs_removexattr(struct dentry *dentry, const char *name)
460 struct inode *inode = dentry->d_inode;
461 struct inode *delegated_inode = NULL;
466 error = __vfs_removexattr_locked(dentry, name, &delegated_inode);
469 if (delegated_inode) {
470 error = break_deleg_wait(&delegated_inode);
477 EXPORT_SYMBOL_GPL(vfs_removexattr);
480 * Extended attribute SET operations
483 setxattr(struct dentry *d, const char __user *name, const void __user *value,
484 size_t size, int flags)
488 char kname[XATTR_NAME_MAX + 1];
490 if (flags & ~(XATTR_CREATE|XATTR_REPLACE))
493 error = strncpy_from_user(kname, name, sizeof(kname));
494 if (error == 0 || error == sizeof(kname))
500 if (size > XATTR_SIZE_MAX)
502 kvalue = kvmalloc(size, GFP_KERNEL);
505 if (copy_from_user(kvalue, value, size)) {
509 if ((strcmp(kname, XATTR_NAME_POSIX_ACL_ACCESS) == 0) ||
510 (strcmp(kname, XATTR_NAME_POSIX_ACL_DEFAULT) == 0))
511 posix_acl_fix_xattr_from_user(kvalue, size);
512 else if (strcmp(kname, XATTR_NAME_CAPS) == 0) {
513 error = cap_convert_nscap(d, &kvalue, size);
520 error = vfs_setxattr(d, kname, kvalue, size, flags);
527 static int path_setxattr(const char __user *pathname,
528 const char __user *name, const void __user *value,
529 size_t size, int flags, unsigned int lookup_flags)
534 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
537 error = mnt_want_write(path.mnt);
539 error = setxattr(path.dentry, name, value, size, flags);
540 mnt_drop_write(path.mnt);
543 if (retry_estale(error, lookup_flags)) {
544 lookup_flags |= LOOKUP_REVAL;
550 SYSCALL_DEFINE5(setxattr, const char __user *, pathname,
551 const char __user *, name, const void __user *, value,
552 size_t, size, int, flags)
554 return path_setxattr(pathname, name, value, size, flags, LOOKUP_FOLLOW);
557 SYSCALL_DEFINE5(lsetxattr, const char __user *, pathname,
558 const char __user *, name, const void __user *, value,
559 size_t, size, int, flags)
561 return path_setxattr(pathname, name, value, size, flags, 0);
564 SYSCALL_DEFINE5(fsetxattr, int, fd, const char __user *, name,
565 const void __user *,value, size_t, size, int, flags)
567 struct fd f = fdget(fd);
573 error = mnt_want_write_file(f.file);
575 error = setxattr(f.file->f_path.dentry, name, value, size, flags);
576 mnt_drop_write_file(f.file);
583 * Extended attribute GET operations
586 getxattr(struct dentry *d, const char __user *name, void __user *value,
591 char kname[XATTR_NAME_MAX + 1];
593 error = strncpy_from_user(kname, name, sizeof(kname));
594 if (error == 0 || error == sizeof(kname))
600 if (size > XATTR_SIZE_MAX)
601 size = XATTR_SIZE_MAX;
602 kvalue = kvzalloc(size, GFP_KERNEL);
607 error = vfs_getxattr(d, kname, kvalue, size);
609 if ((strcmp(kname, XATTR_NAME_POSIX_ACL_ACCESS) == 0) ||
610 (strcmp(kname, XATTR_NAME_POSIX_ACL_DEFAULT) == 0))
611 posix_acl_fix_xattr_to_user(kvalue, error);
612 if (size && copy_to_user(value, kvalue, error))
614 } else if (error == -ERANGE && size >= XATTR_SIZE_MAX) {
615 /* The file system tried to returned a value bigger
616 than XATTR_SIZE_MAX bytes. Not possible. */
625 static ssize_t path_getxattr(const char __user *pathname,
626 const char __user *name, void __user *value,
627 size_t size, unsigned int lookup_flags)
632 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
635 error = getxattr(path.dentry, name, value, size);
637 if (retry_estale(error, lookup_flags)) {
638 lookup_flags |= LOOKUP_REVAL;
644 SYSCALL_DEFINE4(getxattr, const char __user *, pathname,
645 const char __user *, name, void __user *, value, size_t, size)
647 return path_getxattr(pathname, name, value, size, LOOKUP_FOLLOW);
650 SYSCALL_DEFINE4(lgetxattr, const char __user *, pathname,
651 const char __user *, name, void __user *, value, size_t, size)
653 return path_getxattr(pathname, name, value, size, 0);
656 SYSCALL_DEFINE4(fgetxattr, int, fd, const char __user *, name,
657 void __user *, value, size_t, size)
659 struct fd f = fdget(fd);
660 ssize_t error = -EBADF;
665 error = getxattr(f.file->f_path.dentry, name, value, size);
671 * Extended attribute LIST operations
674 listxattr(struct dentry *d, char __user *list, size_t size)
680 if (size > XATTR_LIST_MAX)
681 size = XATTR_LIST_MAX;
682 klist = kvmalloc(size, GFP_KERNEL);
687 error = vfs_listxattr(d, klist, size);
689 if (size && copy_to_user(list, klist, error))
691 } else if (error == -ERANGE && size >= XATTR_LIST_MAX) {
692 /* The file system tried to returned a list bigger
693 than XATTR_LIST_MAX bytes. Not possible. */
702 static ssize_t path_listxattr(const char __user *pathname, char __user *list,
703 size_t size, unsigned int lookup_flags)
708 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
711 error = listxattr(path.dentry, list, size);
713 if (retry_estale(error, lookup_flags)) {
714 lookup_flags |= LOOKUP_REVAL;
720 SYSCALL_DEFINE3(listxattr, const char __user *, pathname, char __user *, list,
723 return path_listxattr(pathname, list, size, LOOKUP_FOLLOW);
726 SYSCALL_DEFINE3(llistxattr, const char __user *, pathname, char __user *, list,
729 return path_listxattr(pathname, list, size, 0);
732 SYSCALL_DEFINE3(flistxattr, int, fd, char __user *, list, size_t, size)
734 struct fd f = fdget(fd);
735 ssize_t error = -EBADF;
740 error = listxattr(f.file->f_path.dentry, list, size);
746 * Extended attribute REMOVE operations
749 removexattr(struct dentry *d, const char __user *name)
752 char kname[XATTR_NAME_MAX + 1];
754 error = strncpy_from_user(kname, name, sizeof(kname));
755 if (error == 0 || error == sizeof(kname))
760 return vfs_removexattr(d, kname);
763 static int path_removexattr(const char __user *pathname,
764 const char __user *name, unsigned int lookup_flags)
769 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
772 error = mnt_want_write(path.mnt);
774 error = removexattr(path.dentry, name);
775 mnt_drop_write(path.mnt);
778 if (retry_estale(error, lookup_flags)) {
779 lookup_flags |= LOOKUP_REVAL;
785 SYSCALL_DEFINE2(removexattr, const char __user *, pathname,
786 const char __user *, name)
788 return path_removexattr(pathname, name, LOOKUP_FOLLOW);
791 SYSCALL_DEFINE2(lremovexattr, const char __user *, pathname,
792 const char __user *, name)
794 return path_removexattr(pathname, name, 0);
797 SYSCALL_DEFINE2(fremovexattr, int, fd, const char __user *, name)
799 struct fd f = fdget(fd);
805 error = mnt_want_write_file(f.file);
807 error = removexattr(f.file->f_path.dentry, name);
808 mnt_drop_write_file(f.file);
815 * Combine the results of the list() operation from every xattr_handler in the
819 generic_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size)
821 const struct xattr_handler *handler, **handlers = dentry->d_sb->s_xattr;
822 unsigned int size = 0;
825 for_each_xattr_handler(handlers, handler) {
826 if (!handler->name ||
827 (handler->list && !handler->list(dentry)))
829 size += strlen(handler->name) + 1;
835 for_each_xattr_handler(handlers, handler) {
836 if (!handler->name ||
837 (handler->list && !handler->list(dentry)))
839 len = strlen(handler->name);
840 if (len + 1 > buffer_size)
842 memcpy(buf, handler->name, len + 1);
844 buffer_size -= len + 1;
850 EXPORT_SYMBOL(generic_listxattr);
853 * xattr_full_name - Compute full attribute name from suffix
855 * @handler: handler of the xattr_handler operation
856 * @name: name passed to the xattr_handler operation
858 * The get and set xattr handler operations are called with the remainder of
859 * the attribute name after skipping the handler's prefix: for example, "foo"
860 * is passed to the get operation of a handler with prefix "user." to get
861 * attribute "user.foo". The full name is still "there" in the name though.
863 * Note: the list xattr handler operation when called from the vfs is passed a
864 * NULL name; some file systems use this operation internally, with varying
867 const char *xattr_full_name(const struct xattr_handler *handler,
870 size_t prefix_len = strlen(xattr_prefix(handler));
872 return name - prefix_len;
874 EXPORT_SYMBOL(xattr_full_name);
877 * Allocate new xattr and copy in the value; but leave the name to callers.
879 struct simple_xattr *simple_xattr_alloc(const void *value, size_t size)
881 struct simple_xattr *new_xattr;
885 len = sizeof(*new_xattr) + size;
886 if (len < sizeof(*new_xattr))
889 new_xattr = kmalloc(len, GFP_KERNEL);
893 new_xattr->size = size;
894 memcpy(new_xattr->value, value, size);
899 * xattr GET operation for in-memory/pseudo filesystems
901 int simple_xattr_get(struct simple_xattrs *xattrs, const char *name,
902 void *buffer, size_t size)
904 struct simple_xattr *xattr;
907 spin_lock(&xattrs->lock);
908 list_for_each_entry(xattr, &xattrs->head, list) {
909 if (strcmp(name, xattr->name))
914 if (size < xattr->size)
917 memcpy(buffer, xattr->value, xattr->size);
921 spin_unlock(&xattrs->lock);
926 * simple_xattr_set - xattr SET operation for in-memory/pseudo filesystems
927 * @xattrs: target simple_xattr list
928 * @name: name of the extended attribute
929 * @value: value of the xattr. If %NULL, will remove the attribute.
930 * @size: size of the new xattr
931 * @flags: %XATTR_{CREATE|REPLACE}
933 * %XATTR_CREATE is set, the xattr shouldn't exist already; otherwise fails
934 * with -EEXIST. If %XATTR_REPLACE is set, the xattr should exist;
935 * otherwise, fails with -ENODATA.
937 * Returns 0 on success, -errno on failure.
939 int simple_xattr_set(struct simple_xattrs *xattrs, const char *name,
940 const void *value, size_t size, int flags)
942 struct simple_xattr *xattr;
943 struct simple_xattr *new_xattr = NULL;
946 /* value == NULL means remove */
948 new_xattr = simple_xattr_alloc(value, size);
952 new_xattr->name = kstrdup(name, GFP_KERNEL);
953 if (!new_xattr->name) {
959 spin_lock(&xattrs->lock);
960 list_for_each_entry(xattr, &xattrs->head, list) {
961 if (!strcmp(name, xattr->name)) {
962 if (flags & XATTR_CREATE) {
965 } else if (new_xattr) {
966 list_replace(&xattr->list, &new_xattr->list);
968 list_del(&xattr->list);
973 if (flags & XATTR_REPLACE) {
977 list_add(&new_xattr->list, &xattrs->head);
981 spin_unlock(&xattrs->lock);
990 static bool xattr_is_trusted(const char *name)
992 return !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN);
995 static int xattr_list_one(char **buffer, ssize_t *remaining_size,
998 size_t len = strlen(name) + 1;
1000 if (*remaining_size < len)
1002 memcpy(*buffer, name, len);
1005 *remaining_size -= len;
1010 * xattr LIST operation for in-memory/pseudo filesystems
1012 ssize_t simple_xattr_list(struct inode *inode, struct simple_xattrs *xattrs,
1013 char *buffer, size_t size)
1015 bool trusted = capable(CAP_SYS_ADMIN);
1016 struct simple_xattr *xattr;
1017 ssize_t remaining_size = size;
1020 #ifdef CONFIG_FS_POSIX_ACL
1021 if (IS_POSIXACL(inode)) {
1023 err = xattr_list_one(&buffer, &remaining_size,
1024 XATTR_NAME_POSIX_ACL_ACCESS);
1028 if (inode->i_default_acl) {
1029 err = xattr_list_one(&buffer, &remaining_size,
1030 XATTR_NAME_POSIX_ACL_DEFAULT);
1037 spin_lock(&xattrs->lock);
1038 list_for_each_entry(xattr, &xattrs->head, list) {
1039 /* skip "trusted." attributes for unprivileged callers */
1040 if (!trusted && xattr_is_trusted(xattr->name))
1043 err = xattr_list_one(&buffer, &remaining_size, xattr->name);
1047 spin_unlock(&xattrs->lock);
1049 return err ? err : size - remaining_size;
1053 * Adds an extended attribute to the list
1055 void simple_xattr_list_add(struct simple_xattrs *xattrs,
1056 struct simple_xattr *new_xattr)
1058 spin_lock(&xattrs->lock);
1059 list_add(&new_xattr->list, &xattrs->head);
1060 spin_unlock(&xattrs->lock);