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>
29 strcmp_prefix(const char *a, const char *a_prefix)
31 while (*a_prefix && *a == *a_prefix) {
35 return *a_prefix ? NULL : a;
39 * In order to implement different sets of xattr operations for each xattr
40 * prefix, a filesystem should create a null-terminated array of struct
41 * xattr_handler (one for each prefix) and hang a pointer to it off of the
42 * s_xattr field of the superblock.
44 #define for_each_xattr_handler(handlers, handler) \
46 for ((handler) = *(handlers)++; \
48 (handler) = *(handlers)++)
51 * Find the xattr_handler with the matching prefix.
53 static const struct xattr_handler *
54 xattr_resolve_name(struct inode *inode, const char **name)
56 const struct xattr_handler **handlers = inode->i_sb->s_xattr;
57 const struct xattr_handler *handler;
59 if (!(inode->i_opflags & IOP_XATTR)) {
60 if (unlikely(is_bad_inode(inode)))
62 return ERR_PTR(-EOPNOTSUPP);
64 for_each_xattr_handler(handlers, handler) {
67 n = strcmp_prefix(*name, xattr_prefix(handler));
69 if (!handler->prefix ^ !*n) {
72 return ERR_PTR(-EINVAL);
78 return ERR_PTR(-EOPNOTSUPP);
82 * Check permissions for extended attribute access. This is a bit complicated
83 * because different namespaces have very different rules.
86 xattr_permission(struct inode *inode, const char *name, int mask)
89 * We can never set or remove an extended attribute on a read-only
90 * filesystem or on an immutable / append-only inode.
92 if (mask & MAY_WRITE) {
93 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
96 * Updating an xattr will likely cause i_uid and i_gid
97 * to be writen back improperly if their true value is
100 if (HAS_UNMAPPED_ID(inode))
105 * No restriction for security.* and system.* from the VFS. Decision
106 * on these is left to the underlying filesystem / security module.
108 if (!strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN) ||
109 !strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
113 * The trusted.* namespace can only be accessed by privileged users.
115 if (!strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN)) {
116 if (!capable(CAP_SYS_ADMIN))
117 return (mask & MAY_WRITE) ? -EPERM : -ENODATA;
122 * In the user.* namespace, only regular files and directories can have
123 * extended attributes. For sticky directories, only the owner and
124 * privileged users can write attributes.
126 if (!strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN)) {
127 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
128 return (mask & MAY_WRITE) ? -EPERM : -ENODATA;
129 if (S_ISDIR(inode->i_mode) && (inode->i_mode & S_ISVTX) &&
130 (mask & MAY_WRITE) && !inode_owner_or_capable(inode))
134 return inode_permission(inode, mask);
138 __vfs_setxattr(struct dentry *dentry, struct inode *inode, const char *name,
139 const void *value, size_t size, int flags)
141 const struct xattr_handler *handler;
143 handler = xattr_resolve_name(inode, &name);
145 return PTR_ERR(handler);
149 value = ""; /* empty EA, do not remove */
150 return handler->set(handler, dentry, inode, name, value, size, flags);
152 EXPORT_SYMBOL(__vfs_setxattr);
155 * __vfs_setxattr_noperm - perform setxattr operation without performing
158 * @dentry - object to perform setxattr on
159 * @name - xattr name to set
160 * @value - value to set @name to
161 * @size - size of @value
162 * @flags - flags to pass into filesystem operations
164 * returns the result of the internal setxattr or setsecurity operations.
166 * This function requires the caller to lock the inode's i_mutex before it
167 * is executed. It also assumes that the caller will make the appropriate
170 int __vfs_setxattr_noperm(struct dentry *dentry, const char *name,
171 const void *value, size_t size, int flags)
173 struct inode *inode = dentry->d_inode;
175 int issec = !strncmp(name, XATTR_SECURITY_PREFIX,
176 XATTR_SECURITY_PREFIX_LEN);
179 inode->i_flags &= ~S_NOSEC;
180 if (inode->i_opflags & IOP_XATTR) {
181 error = __vfs_setxattr(dentry, inode, name, value, size, flags);
183 fsnotify_xattr(dentry);
184 security_inode_post_setxattr(dentry, name, value,
188 if (unlikely(is_bad_inode(inode)))
191 if (error == -EAGAIN) {
195 const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
197 error = security_inode_setsecurity(inode, suffix, value,
200 fsnotify_xattr(dentry);
208 * __vfs_setxattr_locked: set an extended attribute while holding the inode
211 * @dentry - object to perform setxattr on
212 * @name - xattr name to set
213 * @value - value to set @name to
214 * @size - size of @value
215 * @flags - flags to pass into filesystem operations
216 * @delegated_inode - on return, will contain an inode pointer that
217 * a delegation was broken on, NULL if none.
220 __vfs_setxattr_locked(struct dentry *dentry, const char *name,
221 const void *value, size_t size, int flags,
222 struct inode **delegated_inode)
224 struct inode *inode = dentry->d_inode;
227 error = xattr_permission(inode, name, MAY_WRITE);
231 error = security_inode_setxattr(dentry, name, value, size, flags);
235 error = try_break_deleg(inode, delegated_inode);
239 error = __vfs_setxattr_noperm(dentry, name, value, size, flags);
244 EXPORT_SYMBOL_GPL(__vfs_setxattr_locked);
247 vfs_setxattr(struct dentry *dentry, const char *name, const void *value,
248 size_t size, int flags)
250 struct inode *inode = dentry->d_inode;
251 struct inode *delegated_inode = NULL;
256 error = __vfs_setxattr_locked(dentry, name, value, size, flags,
260 if (delegated_inode) {
261 error = break_deleg_wait(&delegated_inode);
267 EXPORT_SYMBOL_GPL(vfs_setxattr);
270 xattr_getsecurity(struct inode *inode, const char *name, void *value,
276 if (!value || !size) {
277 len = security_inode_getsecurity(inode, name, &buffer, false);
281 len = security_inode_getsecurity(inode, name, &buffer, true);
288 memcpy(value, buffer, len);
294 EXPORT_SYMBOL_GPL(xattr_getsecurity);
297 * vfs_getxattr_alloc - allocate memory, if necessary, before calling getxattr
299 * Allocate memory, if not already allocated, or re-allocate correct size,
300 * before retrieving the extended attribute.
302 * Returns the result of alloc, if failed, or the getxattr operation.
305 vfs_getxattr_alloc(struct dentry *dentry, const char *name, char **xattr_value,
306 size_t xattr_size, gfp_t flags)
308 const struct xattr_handler *handler;
309 struct inode *inode = dentry->d_inode;
310 char *value = *xattr_value;
313 error = xattr_permission(inode, name, MAY_READ);
317 handler = xattr_resolve_name(inode, &name);
319 return PTR_ERR(handler);
322 error = handler->get(handler, dentry, inode, name, NULL, 0);
326 if (!value || (error > xattr_size)) {
327 value = krealloc(*xattr_value, error + 1, flags);
330 memset(value, 0, error + 1);
333 error = handler->get(handler, dentry, inode, name, value, error);
334 *xattr_value = value;
339 __vfs_getxattr(struct dentry *dentry, struct inode *inode, const char *name,
340 void *value, size_t size)
342 const struct xattr_handler *handler;
344 handler = xattr_resolve_name(inode, &name);
346 return PTR_ERR(handler);
349 return handler->get(handler, dentry, inode, name, value, size);
351 EXPORT_SYMBOL(__vfs_getxattr);
354 vfs_getxattr(struct dentry *dentry, const char *name, void *value, size_t size)
356 struct inode *inode = dentry->d_inode;
359 error = xattr_permission(inode, name, MAY_READ);
363 error = security_inode_getxattr(dentry, name);
367 if (!strncmp(name, XATTR_SECURITY_PREFIX,
368 XATTR_SECURITY_PREFIX_LEN)) {
369 const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
370 int ret = xattr_getsecurity(inode, suffix, value, size);
372 * Only overwrite the return value if a security module
373 * is actually active.
375 if (ret == -EOPNOTSUPP)
380 return __vfs_getxattr(dentry, inode, name, value, size);
382 EXPORT_SYMBOL_GPL(vfs_getxattr);
385 vfs_listxattr(struct dentry *dentry, char *list, size_t size)
387 struct inode *inode = d_inode(dentry);
390 error = security_inode_listxattr(dentry);
393 if (inode->i_op->listxattr && (inode->i_opflags & IOP_XATTR)) {
395 error = inode->i_op->listxattr(dentry, list, size);
397 error = security_inode_listsecurity(inode, list, size);
398 if (size && error > size)
403 EXPORT_SYMBOL_GPL(vfs_listxattr);
406 __vfs_removexattr(struct dentry *dentry, const char *name)
408 struct inode *inode = d_inode(dentry);
409 const struct xattr_handler *handler;
411 handler = xattr_resolve_name(inode, &name);
413 return PTR_ERR(handler);
416 return handler->set(handler, dentry, inode, name, NULL, 0, XATTR_REPLACE);
418 EXPORT_SYMBOL(__vfs_removexattr);
421 * __vfs_removexattr_locked: set an extended attribute while holding the inode
424 * @dentry - object to perform setxattr on
425 * @name - name of xattr to remove
426 * @delegated_inode - on return, will contain an inode pointer that
427 * a delegation was broken on, NULL if none.
430 __vfs_removexattr_locked(struct dentry *dentry, const char *name,
431 struct inode **delegated_inode)
433 struct inode *inode = dentry->d_inode;
436 error = xattr_permission(inode, name, MAY_WRITE);
440 error = security_inode_removexattr(dentry, name);
444 error = try_break_deleg(inode, delegated_inode);
448 error = __vfs_removexattr(dentry, name);
451 fsnotify_xattr(dentry);
452 evm_inode_post_removexattr(dentry, name);
458 EXPORT_SYMBOL_GPL(__vfs_removexattr_locked);
461 vfs_removexattr(struct dentry *dentry, const char *name)
463 struct inode *inode = dentry->d_inode;
464 struct inode *delegated_inode = NULL;
469 error = __vfs_removexattr_locked(dentry, name, &delegated_inode);
472 if (delegated_inode) {
473 error = break_deleg_wait(&delegated_inode);
480 EXPORT_SYMBOL_GPL(vfs_removexattr);
483 * Extended attribute SET operations
486 setxattr(struct dentry *d, const char __user *name, const void __user *value,
487 size_t size, int flags)
491 char kname[XATTR_NAME_MAX + 1];
493 if (flags & ~(XATTR_CREATE|XATTR_REPLACE))
496 error = strncpy_from_user(kname, name, sizeof(kname));
497 if (error == 0 || error == sizeof(kname))
503 if (size > XATTR_SIZE_MAX)
505 kvalue = kvmalloc(size, GFP_KERNEL);
508 if (copy_from_user(kvalue, value, size)) {
512 if ((strcmp(kname, XATTR_NAME_POSIX_ACL_ACCESS) == 0) ||
513 (strcmp(kname, XATTR_NAME_POSIX_ACL_DEFAULT) == 0))
514 posix_acl_fix_xattr_from_user(kvalue, size);
515 else if (strcmp(kname, XATTR_NAME_CAPS) == 0) {
516 error = cap_convert_nscap(d, &kvalue, size);
523 error = vfs_setxattr(d, kname, kvalue, size, flags);
530 static int path_setxattr(const char __user *pathname,
531 const char __user *name, const void __user *value,
532 size_t size, int flags, unsigned int lookup_flags)
537 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
540 error = mnt_want_write(path.mnt);
542 error = setxattr(path.dentry, name, value, size, flags);
543 mnt_drop_write(path.mnt);
546 if (retry_estale(error, lookup_flags)) {
547 lookup_flags |= LOOKUP_REVAL;
553 SYSCALL_DEFINE5(setxattr, const char __user *, pathname,
554 const char __user *, name, const void __user *, value,
555 size_t, size, int, flags)
557 return path_setxattr(pathname, name, value, size, flags, LOOKUP_FOLLOW);
560 SYSCALL_DEFINE5(lsetxattr, const char __user *, pathname,
561 const char __user *, name, const void __user *, value,
562 size_t, size, int, flags)
564 return path_setxattr(pathname, name, value, size, flags, 0);
567 SYSCALL_DEFINE5(fsetxattr, int, fd, const char __user *, name,
568 const void __user *,value, size_t, size, int, flags)
570 struct fd f = fdget(fd);
576 error = mnt_want_write_file_path(f.file);
578 error = setxattr(f.file->f_path.dentry, name, value, size, flags);
579 mnt_drop_write_file_path(f.file);
586 * Extended attribute GET operations
589 getxattr(struct dentry *d, const char __user *name, void __user *value,
594 char kname[XATTR_NAME_MAX + 1];
596 error = strncpy_from_user(kname, name, sizeof(kname));
597 if (error == 0 || error == sizeof(kname))
603 if (size > XATTR_SIZE_MAX)
604 size = XATTR_SIZE_MAX;
605 kvalue = kvzalloc(size, GFP_KERNEL);
610 error = vfs_getxattr(d, kname, kvalue, size);
612 if ((strcmp(kname, XATTR_NAME_POSIX_ACL_ACCESS) == 0) ||
613 (strcmp(kname, XATTR_NAME_POSIX_ACL_DEFAULT) == 0))
614 posix_acl_fix_xattr_to_user(kvalue, error);
615 if (size && copy_to_user(value, kvalue, error))
617 } else if (error == -ERANGE && size >= XATTR_SIZE_MAX) {
618 /* The file system tried to returned a value bigger
619 than XATTR_SIZE_MAX bytes. Not possible. */
628 static ssize_t path_getxattr(const char __user *pathname,
629 const char __user *name, void __user *value,
630 size_t size, unsigned int lookup_flags)
635 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
638 error = getxattr(path.dentry, name, value, size);
640 if (retry_estale(error, lookup_flags)) {
641 lookup_flags |= LOOKUP_REVAL;
647 SYSCALL_DEFINE4(getxattr, const char __user *, pathname,
648 const char __user *, name, void __user *, value, size_t, size)
650 return path_getxattr(pathname, name, value, size, LOOKUP_FOLLOW);
653 SYSCALL_DEFINE4(lgetxattr, const char __user *, pathname,
654 const char __user *, name, void __user *, value, size_t, size)
656 return path_getxattr(pathname, name, value, size, 0);
659 SYSCALL_DEFINE4(fgetxattr, int, fd, const char __user *, name,
660 void __user *, value, size_t, size)
662 struct fd f = fdget(fd);
663 ssize_t error = -EBADF;
668 error = getxattr(f.file->f_path.dentry, name, value, size);
674 * Extended attribute LIST operations
677 listxattr(struct dentry *d, char __user *list, size_t size)
683 if (size > XATTR_LIST_MAX)
684 size = XATTR_LIST_MAX;
685 klist = kvmalloc(size, GFP_KERNEL);
690 error = vfs_listxattr(d, klist, size);
692 if (size && copy_to_user(list, klist, error))
694 } else if (error == -ERANGE && size >= XATTR_LIST_MAX) {
695 /* The file system tried to returned a list bigger
696 than XATTR_LIST_MAX bytes. Not possible. */
705 static ssize_t path_listxattr(const char __user *pathname, char __user *list,
706 size_t size, unsigned int lookup_flags)
711 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
714 error = listxattr(path.dentry, list, size);
716 if (retry_estale(error, lookup_flags)) {
717 lookup_flags |= LOOKUP_REVAL;
723 SYSCALL_DEFINE3(listxattr, const char __user *, pathname, char __user *, list,
726 return path_listxattr(pathname, list, size, LOOKUP_FOLLOW);
729 SYSCALL_DEFINE3(llistxattr, const char __user *, pathname, char __user *, list,
732 return path_listxattr(pathname, list, size, 0);
735 SYSCALL_DEFINE3(flistxattr, int, fd, char __user *, list, size_t, size)
737 struct fd f = fdget(fd);
738 ssize_t error = -EBADF;
743 error = listxattr(f.file->f_path.dentry, list, size);
749 * Extended attribute REMOVE operations
752 removexattr(struct dentry *d, const char __user *name)
755 char kname[XATTR_NAME_MAX + 1];
757 error = strncpy_from_user(kname, name, sizeof(kname));
758 if (error == 0 || error == sizeof(kname))
763 return vfs_removexattr(d, kname);
766 static int path_removexattr(const char __user *pathname,
767 const char __user *name, unsigned int lookup_flags)
772 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
775 error = mnt_want_write(path.mnt);
777 error = removexattr(path.dentry, name);
778 mnt_drop_write(path.mnt);
781 if (retry_estale(error, lookup_flags)) {
782 lookup_flags |= LOOKUP_REVAL;
788 SYSCALL_DEFINE2(removexattr, const char __user *, pathname,
789 const char __user *, name)
791 return path_removexattr(pathname, name, LOOKUP_FOLLOW);
794 SYSCALL_DEFINE2(lremovexattr, const char __user *, pathname,
795 const char __user *, name)
797 return path_removexattr(pathname, name, 0);
800 SYSCALL_DEFINE2(fremovexattr, int, fd, const char __user *, name)
802 struct fd f = fdget(fd);
808 error = mnt_want_write_file_path(f.file);
810 error = removexattr(f.file->f_path.dentry, name);
811 mnt_drop_write_file_path(f.file);
818 * Combine the results of the list() operation from every xattr_handler in the
822 generic_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size)
824 const struct xattr_handler *handler, **handlers = dentry->d_sb->s_xattr;
825 unsigned int size = 0;
828 for_each_xattr_handler(handlers, handler) {
829 if (!handler->name ||
830 (handler->list && !handler->list(dentry)))
832 size += strlen(handler->name) + 1;
838 for_each_xattr_handler(handlers, handler) {
839 if (!handler->name ||
840 (handler->list && !handler->list(dentry)))
842 len = strlen(handler->name);
843 if (len + 1 > buffer_size)
845 memcpy(buf, handler->name, len + 1);
847 buffer_size -= len + 1;
853 EXPORT_SYMBOL(generic_listxattr);
856 * xattr_full_name - Compute full attribute name from suffix
858 * @handler: handler of the xattr_handler operation
859 * @name: name passed to the xattr_handler operation
861 * The get and set xattr handler operations are called with the remainder of
862 * the attribute name after skipping the handler's prefix: for example, "foo"
863 * is passed to the get operation of a handler with prefix "user." to get
864 * attribute "user.foo". The full name is still "there" in the name though.
866 * Note: the list xattr handler operation when called from the vfs is passed a
867 * NULL name; some file systems use this operation internally, with varying
870 const char *xattr_full_name(const struct xattr_handler *handler,
873 size_t prefix_len = strlen(xattr_prefix(handler));
875 return name - prefix_len;
877 EXPORT_SYMBOL(xattr_full_name);
880 * Allocate new xattr and copy in the value; but leave the name to callers.
882 struct simple_xattr *simple_xattr_alloc(const void *value, size_t size)
884 struct simple_xattr *new_xattr;
888 len = sizeof(*new_xattr) + size;
889 if (len < sizeof(*new_xattr))
892 new_xattr = kmalloc(len, GFP_KERNEL);
896 new_xattr->size = size;
897 memcpy(new_xattr->value, value, size);
902 * xattr GET operation for in-memory/pseudo filesystems
904 int simple_xattr_get(struct simple_xattrs *xattrs, const char *name,
905 void *buffer, size_t size)
907 struct simple_xattr *xattr;
910 spin_lock(&xattrs->lock);
911 list_for_each_entry(xattr, &xattrs->head, list) {
912 if (strcmp(name, xattr->name))
917 if (size < xattr->size)
920 memcpy(buffer, xattr->value, xattr->size);
924 spin_unlock(&xattrs->lock);
929 * simple_xattr_set - xattr SET operation for in-memory/pseudo filesystems
930 * @xattrs: target simple_xattr list
931 * @name: name of the extended attribute
932 * @value: value of the xattr. If %NULL, will remove the attribute.
933 * @size: size of the new xattr
934 * @flags: %XATTR_{CREATE|REPLACE}
936 * %XATTR_CREATE is set, the xattr shouldn't exist already; otherwise fails
937 * with -EEXIST. If %XATTR_REPLACE is set, the xattr should exist;
938 * otherwise, fails with -ENODATA.
940 * Returns 0 on success, -errno on failure.
942 int simple_xattr_set(struct simple_xattrs *xattrs, const char *name,
943 const void *value, size_t size, int flags)
945 struct simple_xattr *xattr;
946 struct simple_xattr *new_xattr = NULL;
949 /* value == NULL means remove */
951 new_xattr = simple_xattr_alloc(value, size);
955 new_xattr->name = kstrdup(name, GFP_KERNEL);
956 if (!new_xattr->name) {
962 spin_lock(&xattrs->lock);
963 list_for_each_entry(xattr, &xattrs->head, list) {
964 if (!strcmp(name, xattr->name)) {
965 if (flags & XATTR_CREATE) {
968 } else if (new_xattr) {
969 list_replace(&xattr->list, &new_xattr->list);
971 list_del(&xattr->list);
976 if (flags & XATTR_REPLACE) {
980 list_add(&new_xattr->list, &xattrs->head);
984 spin_unlock(&xattrs->lock);
993 static bool xattr_is_trusted(const char *name)
995 return !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN);
998 static int xattr_list_one(char **buffer, ssize_t *remaining_size,
1001 size_t len = strlen(name) + 1;
1003 if (*remaining_size < len)
1005 memcpy(*buffer, name, len);
1008 *remaining_size -= len;
1013 * xattr LIST operation for in-memory/pseudo filesystems
1015 ssize_t simple_xattr_list(struct inode *inode, struct simple_xattrs *xattrs,
1016 char *buffer, size_t size)
1018 bool trusted = ns_capable_noaudit(&init_user_ns, CAP_SYS_ADMIN);
1019 struct simple_xattr *xattr;
1020 ssize_t remaining_size = size;
1023 #ifdef CONFIG_FS_POSIX_ACL
1024 if (IS_POSIXACL(inode)) {
1026 err = xattr_list_one(&buffer, &remaining_size,
1027 XATTR_NAME_POSIX_ACL_ACCESS);
1031 if (inode->i_default_acl) {
1032 err = xattr_list_one(&buffer, &remaining_size,
1033 XATTR_NAME_POSIX_ACL_DEFAULT);
1040 spin_lock(&xattrs->lock);
1041 list_for_each_entry(xattr, &xattrs->head, list) {
1042 /* skip "trusted." attributes for unprivileged callers */
1043 if (!trusted && xattr_is_trusted(xattr->name))
1046 err = xattr_list_one(&buffer, &remaining_size, xattr->name);
1050 spin_unlock(&xattrs->lock);
1052 return err ? err : size - remaining_size;
1056 * Adds an extended attribute to the list
1058 void simple_xattr_list_add(struct simple_xattrs *xattrs,
1059 struct simple_xattr *new_xattr)
1061 spin_lock(&xattrs->lock);
1062 list_add(&new_xattr->list, &xattrs->head);
1063 spin_unlock(&xattrs->lock);