1 // SPDX-License-Identifier: GPL-2.0-only
5 Extended attribute handling.
12 #include <linux/filelock.h>
13 #include <linux/slab.h>
14 #include <linux/file.h>
15 #include <linux/xattr.h>
16 #include <linux/mount.h>
17 #include <linux/namei.h>
18 #include <linux/security.h>
19 #include <linux/syscalls.h>
20 #include <linux/export.h>
21 #include <linux/fsnotify.h>
22 #include <linux/audit.h>
23 #include <linux/vmalloc.h>
24 #include <linux/posix_acl_xattr.h>
26 #include <linux/uaccess.h>
31 strcmp_prefix(const char *a, const char *a_prefix)
33 while (*a_prefix && *a == *a_prefix) {
37 return *a_prefix ? NULL : a;
41 * In order to implement different sets of xattr operations for each xattr
42 * prefix, a filesystem should create a null-terminated array of struct
43 * xattr_handler (one for each prefix) and hang a pointer to it off of the
44 * s_xattr field of the superblock.
46 #define for_each_xattr_handler(handlers, handler) \
48 for ((handler) = *(handlers)++; \
50 (handler) = *(handlers)++)
53 * Find the xattr_handler with the matching prefix.
55 static const struct xattr_handler *
56 xattr_resolve_name(struct inode *inode, const char **name)
58 const struct xattr_handler * const *handlers = inode->i_sb->s_xattr;
59 const struct xattr_handler *handler;
61 if (!(inode->i_opflags & IOP_XATTR)) {
62 if (unlikely(is_bad_inode(inode)))
64 return ERR_PTR(-EOPNOTSUPP);
66 for_each_xattr_handler(handlers, handler) {
69 n = strcmp_prefix(*name, xattr_prefix(handler));
71 if (!handler->prefix ^ !*n) {
74 return ERR_PTR(-EINVAL);
80 return ERR_PTR(-EOPNOTSUPP);
84 * may_write_xattr - check whether inode allows writing xattr
85 * @idmap: idmap of the mount the inode was found from
86 * @inode: the inode on which to set an xattr
88 * Check whether the inode allows writing xattrs. Specifically, we can never
89 * set or remove an extended attribute on a read-only filesystem or on an
90 * immutable / append-only inode.
92 * We also need to ensure that the inode has a mapping in the mount to
93 * not risk writing back invalid i_{g,u}id values.
95 * Return: On success zero is returned. On error a negative errno is returned.
97 int may_write_xattr(struct mnt_idmap *idmap, struct inode *inode)
99 if (IS_IMMUTABLE(inode))
101 if (IS_APPEND(inode))
103 if (HAS_UNMAPPED_ID(idmap, inode))
109 * Check permissions for extended attribute access. This is a bit complicated
110 * because different namespaces have very different rules.
113 xattr_permission(struct mnt_idmap *idmap, struct inode *inode,
114 const char *name, int mask)
116 if (mask & MAY_WRITE) {
119 ret = may_write_xattr(idmap, inode);
125 * No restriction for security.* and system.* from the VFS. Decision
126 * on these is left to the underlying filesystem / security module.
128 if (!strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN) ||
129 !strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
133 * The trusted.* namespace can only be accessed by privileged users.
135 if (!strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN)) {
136 if (!capable(CAP_SYS_ADMIN))
137 return (mask & MAY_WRITE) ? -EPERM : -ENODATA;
142 * In the user.* namespace, only regular files and directories can have
143 * extended attributes. For sticky directories, only the owner and
144 * privileged users can write attributes.
146 if (!strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN)) {
147 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
148 return (mask & MAY_WRITE) ? -EPERM : -ENODATA;
149 if (S_ISDIR(inode->i_mode) && (inode->i_mode & S_ISVTX) &&
150 (mask & MAY_WRITE) &&
151 !inode_owner_or_capable(idmap, inode))
155 return inode_permission(idmap, inode, mask);
159 * Look for any handler that deals with the specified namespace.
162 xattr_supports_user_prefix(struct inode *inode)
164 const struct xattr_handler * const *handlers = inode->i_sb->s_xattr;
165 const struct xattr_handler *handler;
167 if (!(inode->i_opflags & IOP_XATTR)) {
168 if (unlikely(is_bad_inode(inode)))
173 for_each_xattr_handler(handlers, handler) {
174 if (!strncmp(xattr_prefix(handler), XATTR_USER_PREFIX,
175 XATTR_USER_PREFIX_LEN))
181 EXPORT_SYMBOL(xattr_supports_user_prefix);
184 __vfs_setxattr(struct mnt_idmap *idmap, struct dentry *dentry,
185 struct inode *inode, const char *name, const void *value,
186 size_t size, int flags)
188 const struct xattr_handler *handler;
190 if (is_posix_acl_xattr(name))
193 handler = xattr_resolve_name(inode, &name);
195 return PTR_ERR(handler);
199 value = ""; /* empty EA, do not remove */
200 return handler->set(handler, idmap, dentry, inode, name, value,
203 EXPORT_SYMBOL(__vfs_setxattr);
206 * __vfs_setxattr_noperm - perform setxattr operation without performing
209 * @idmap: idmap of the mount the inode was found from
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
216 * returns the result of the internal setxattr or setsecurity operations.
218 * This function requires the caller to lock the inode's i_mutex before it
219 * is executed. It also assumes that the caller will make the appropriate
222 int __vfs_setxattr_noperm(struct mnt_idmap *idmap,
223 struct dentry *dentry, const char *name,
224 const void *value, size_t size, int flags)
226 struct inode *inode = dentry->d_inode;
228 int issec = !strncmp(name, XATTR_SECURITY_PREFIX,
229 XATTR_SECURITY_PREFIX_LEN);
232 inode->i_flags &= ~S_NOSEC;
233 if (inode->i_opflags & IOP_XATTR) {
234 error = __vfs_setxattr(idmap, dentry, inode, name, value,
237 fsnotify_xattr(dentry);
238 security_inode_post_setxattr(dentry, name, value,
242 if (unlikely(is_bad_inode(inode)))
245 if (error == -EAGAIN) {
249 const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
251 error = security_inode_setsecurity(inode, suffix, value,
254 fsnotify_xattr(dentry);
262 * __vfs_setxattr_locked - set an extended attribute while holding the inode
265 * @idmap: idmap of the mount of the target inode
266 * @dentry: object to perform setxattr on
267 * @name: xattr name to set
268 * @value: value to set @name to
269 * @size: size of @value
270 * @flags: flags to pass into filesystem operations
271 * @delegated_inode: on return, will contain an inode pointer that
272 * a delegation was broken on, NULL if none.
275 __vfs_setxattr_locked(struct mnt_idmap *idmap, struct dentry *dentry,
276 const char *name, const void *value, size_t size,
277 int flags, struct inode **delegated_inode)
279 struct inode *inode = dentry->d_inode;
282 error = xattr_permission(idmap, inode, name, MAY_WRITE);
286 error = security_inode_setxattr(idmap, dentry, name, value, size,
291 error = try_break_deleg(inode, delegated_inode);
295 error = __vfs_setxattr_noperm(idmap, dentry, name, value,
301 EXPORT_SYMBOL_GPL(__vfs_setxattr_locked);
304 vfs_setxattr(struct mnt_idmap *idmap, struct dentry *dentry,
305 const char *name, const void *value, size_t size, int flags)
307 struct inode *inode = dentry->d_inode;
308 struct inode *delegated_inode = NULL;
309 const void *orig_value = value;
312 if (size && strcmp(name, XATTR_NAME_CAPS) == 0) {
313 error = cap_convert_nscap(idmap, dentry, &value, size);
321 error = __vfs_setxattr_locked(idmap, dentry, name, value, size,
322 flags, &delegated_inode);
325 if (delegated_inode) {
326 error = break_deleg_wait(&delegated_inode);
330 if (value != orig_value)
335 EXPORT_SYMBOL_GPL(vfs_setxattr);
338 xattr_getsecurity(struct mnt_idmap *idmap, struct inode *inode,
339 const char *name, void *value, size_t size)
344 if (!value || !size) {
345 len = security_inode_getsecurity(idmap, inode, name,
350 len = security_inode_getsecurity(idmap, inode, name, &buffer,
358 memcpy(value, buffer, len);
366 * vfs_getxattr_alloc - allocate memory, if necessary, before calling getxattr
368 * Allocate memory, if not already allocated, or re-allocate correct size,
369 * before retrieving the extended attribute. The xattr value buffer should
370 * always be freed by the caller, even on error.
372 * Returns the result of alloc, if failed, or the getxattr operation.
375 vfs_getxattr_alloc(struct mnt_idmap *idmap, struct dentry *dentry,
376 const char *name, char **xattr_value, size_t xattr_size,
379 const struct xattr_handler *handler;
380 struct inode *inode = dentry->d_inode;
381 char *value = *xattr_value;
384 error = xattr_permission(idmap, inode, name, MAY_READ);
388 handler = xattr_resolve_name(inode, &name);
390 return PTR_ERR(handler);
393 error = handler->get(handler, dentry, inode, name, NULL, 0);
397 if (!value || (error > xattr_size)) {
398 value = krealloc(*xattr_value, error + 1, flags);
401 memset(value, 0, error + 1);
404 error = handler->get(handler, dentry, inode, name, value, error);
405 *xattr_value = value;
410 __vfs_getxattr(struct dentry *dentry, struct inode *inode, const char *name,
411 void *value, size_t size)
413 const struct xattr_handler *handler;
415 if (is_posix_acl_xattr(name))
418 handler = xattr_resolve_name(inode, &name);
420 return PTR_ERR(handler);
423 return handler->get(handler, dentry, inode, name, value, size);
425 EXPORT_SYMBOL(__vfs_getxattr);
428 vfs_getxattr(struct mnt_idmap *idmap, struct dentry *dentry,
429 const char *name, void *value, size_t size)
431 struct inode *inode = dentry->d_inode;
434 error = xattr_permission(idmap, inode, name, MAY_READ);
438 error = security_inode_getxattr(dentry, name);
442 if (!strncmp(name, XATTR_SECURITY_PREFIX,
443 XATTR_SECURITY_PREFIX_LEN)) {
444 const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
445 int ret = xattr_getsecurity(idmap, inode, suffix, value,
448 * Only overwrite the return value if a security module
449 * is actually active.
451 if (ret == -EOPNOTSUPP)
456 return __vfs_getxattr(dentry, inode, name, value, size);
458 EXPORT_SYMBOL_GPL(vfs_getxattr);
461 * vfs_listxattr - retrieve \0 separated list of xattr names
462 * @dentry: the dentry from whose inode the xattr names are retrieved
463 * @list: buffer to store xattr names into
464 * @size: size of the buffer
466 * This function returns the names of all xattrs associated with the
469 * Note, for legacy reasons the vfs_listxattr() function lists POSIX
470 * ACLs as well. Since POSIX ACLs are decoupled from IOP_XATTR the
471 * vfs_listxattr() function doesn't check for this flag since a
472 * filesystem could implement POSIX ACLs without implementing any other
475 * However, since all codepaths that remove IOP_XATTR also assign of
476 * inode operations that either don't implement or implement a stub
477 * ->listxattr() operation.
479 * Return: On success, the size of the buffer that was used. On error a
480 * negative error code.
483 vfs_listxattr(struct dentry *dentry, char *list, size_t size)
485 struct inode *inode = d_inode(dentry);
488 error = security_inode_listxattr(dentry);
492 if (inode->i_op->listxattr) {
493 error = inode->i_op->listxattr(dentry, list, size);
495 error = security_inode_listsecurity(inode, list, size);
496 if (size && error > size)
501 EXPORT_SYMBOL_GPL(vfs_listxattr);
504 __vfs_removexattr(struct mnt_idmap *idmap, struct dentry *dentry,
507 struct inode *inode = d_inode(dentry);
508 const struct xattr_handler *handler;
510 if (is_posix_acl_xattr(name))
513 handler = xattr_resolve_name(inode, &name);
515 return PTR_ERR(handler);
518 return handler->set(handler, idmap, dentry, inode, name, NULL, 0,
521 EXPORT_SYMBOL(__vfs_removexattr);
524 * __vfs_removexattr_locked - set an extended attribute while holding the inode
527 * @idmap: idmap of the mount of the target inode
528 * @dentry: object to perform setxattr on
529 * @name: name of xattr to remove
530 * @delegated_inode: on return, will contain an inode pointer that
531 * a delegation was broken on, NULL if none.
534 __vfs_removexattr_locked(struct mnt_idmap *idmap,
535 struct dentry *dentry, const char *name,
536 struct inode **delegated_inode)
538 struct inode *inode = dentry->d_inode;
541 error = xattr_permission(idmap, inode, name, MAY_WRITE);
545 error = security_inode_removexattr(idmap, dentry, name);
549 error = try_break_deleg(inode, delegated_inode);
553 error = __vfs_removexattr(idmap, dentry, name);
557 fsnotify_xattr(dentry);
558 security_inode_post_removexattr(dentry, name);
563 EXPORT_SYMBOL_GPL(__vfs_removexattr_locked);
566 vfs_removexattr(struct mnt_idmap *idmap, struct dentry *dentry,
569 struct inode *inode = dentry->d_inode;
570 struct inode *delegated_inode = NULL;
575 error = __vfs_removexattr_locked(idmap, dentry,
576 name, &delegated_inode);
579 if (delegated_inode) {
580 error = break_deleg_wait(&delegated_inode);
587 EXPORT_SYMBOL_GPL(vfs_removexattr);
590 * Extended attribute SET operations
593 int setxattr_copy(const char __user *name, struct xattr_ctx *ctx)
597 if (ctx->flags & ~(XATTR_CREATE|XATTR_REPLACE))
600 error = strncpy_from_user(ctx->kname->name, name,
601 sizeof(ctx->kname->name));
602 if (error == 0 || error == sizeof(ctx->kname->name))
609 if (ctx->size > XATTR_SIZE_MAX)
612 ctx->kvalue = vmemdup_user(ctx->cvalue, ctx->size);
613 if (IS_ERR(ctx->kvalue)) {
614 error = PTR_ERR(ctx->kvalue);
622 int do_setxattr(struct mnt_idmap *idmap, struct dentry *dentry,
623 struct xattr_ctx *ctx)
625 if (is_posix_acl_xattr(ctx->kname->name))
626 return do_set_acl(idmap, dentry, ctx->kname->name,
627 ctx->kvalue, ctx->size);
629 return vfs_setxattr(idmap, dentry, ctx->kname->name,
630 ctx->kvalue, ctx->size, ctx->flags);
633 static int path_setxattr(const char __user *pathname,
634 const char __user *name, const void __user *value,
635 size_t size, int flags, unsigned int lookup_flags)
637 struct xattr_name kname;
638 struct xattr_ctx ctx = {
648 error = setxattr_copy(name, &ctx);
653 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
656 error = mnt_want_write(path.mnt);
658 error = do_setxattr(mnt_idmap(path.mnt), path.dentry, &ctx);
659 mnt_drop_write(path.mnt);
662 if (retry_estale(error, lookup_flags)) {
663 lookup_flags |= LOOKUP_REVAL;
672 SYSCALL_DEFINE5(setxattr, const char __user *, pathname,
673 const char __user *, name, const void __user *, value,
674 size_t, size, int, flags)
676 return path_setxattr(pathname, name, value, size, flags, LOOKUP_FOLLOW);
679 SYSCALL_DEFINE5(lsetxattr, const char __user *, pathname,
680 const char __user *, name, const void __user *, value,
681 size_t, size, int, flags)
683 return path_setxattr(pathname, name, value, size, flags, 0);
686 SYSCALL_DEFINE5(fsetxattr, int, fd, const char __user *, name,
687 const void __user *,value, size_t, size, int, flags)
689 struct xattr_name kname;
690 struct xattr_ctx ctx = {
704 error = setxattr_copy(name, &ctx);
708 error = mnt_want_write_file(f.file);
710 error = do_setxattr(file_mnt_idmap(f.file),
711 f.file->f_path.dentry, &ctx);
712 mnt_drop_write_file(f.file);
719 * Extended attribute GET operations
722 do_getxattr(struct mnt_idmap *idmap, struct dentry *d,
723 struct xattr_ctx *ctx)
726 char *kname = ctx->kname->name;
729 if (ctx->size > XATTR_SIZE_MAX)
730 ctx->size = XATTR_SIZE_MAX;
731 ctx->kvalue = kvzalloc(ctx->size, GFP_KERNEL);
736 if (is_posix_acl_xattr(ctx->kname->name))
737 error = do_get_acl(idmap, d, kname, ctx->kvalue, ctx->size);
739 error = vfs_getxattr(idmap, d, kname, ctx->kvalue, ctx->size);
741 if (ctx->size && copy_to_user(ctx->value, ctx->kvalue, error))
743 } else if (error == -ERANGE && ctx->size >= XATTR_SIZE_MAX) {
744 /* The file system tried to returned a value bigger
745 than XATTR_SIZE_MAX bytes. Not possible. */
753 getxattr(struct mnt_idmap *idmap, struct dentry *d,
754 const char __user *name, void __user *value, size_t size)
757 struct xattr_name kname;
758 struct xattr_ctx ctx = {
766 error = strncpy_from_user(kname.name, name, sizeof(kname.name));
767 if (error == 0 || error == sizeof(kname.name))
772 error = do_getxattr(idmap, d, &ctx);
778 static ssize_t path_getxattr(const char __user *pathname,
779 const char __user *name, void __user *value,
780 size_t size, unsigned int lookup_flags)
785 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
788 error = getxattr(mnt_idmap(path.mnt), path.dentry, name, value, size);
790 if (retry_estale(error, lookup_flags)) {
791 lookup_flags |= LOOKUP_REVAL;
797 SYSCALL_DEFINE4(getxattr, const char __user *, pathname,
798 const char __user *, name, void __user *, value, size_t, size)
800 return path_getxattr(pathname, name, value, size, LOOKUP_FOLLOW);
803 SYSCALL_DEFINE4(lgetxattr, const char __user *, pathname,
804 const char __user *, name, void __user *, value, size_t, size)
806 return path_getxattr(pathname, name, value, size, 0);
809 SYSCALL_DEFINE4(fgetxattr, int, fd, const char __user *, name,
810 void __user *, value, size_t, size)
812 struct fd f = fdget(fd);
813 ssize_t error = -EBADF;
818 error = getxattr(file_mnt_idmap(f.file), f.file->f_path.dentry,
825 * Extended attribute LIST operations
828 listxattr(struct dentry *d, char __user *list, size_t size)
834 if (size > XATTR_LIST_MAX)
835 size = XATTR_LIST_MAX;
836 klist = kvmalloc(size, GFP_KERNEL);
841 error = vfs_listxattr(d, klist, size);
843 if (size && copy_to_user(list, klist, error))
845 } else if (error == -ERANGE && size >= XATTR_LIST_MAX) {
846 /* The file system tried to returned a list bigger
847 than XATTR_LIST_MAX bytes. Not possible. */
856 static ssize_t path_listxattr(const char __user *pathname, char __user *list,
857 size_t size, unsigned int lookup_flags)
862 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
865 error = listxattr(path.dentry, list, size);
867 if (retry_estale(error, lookup_flags)) {
868 lookup_flags |= LOOKUP_REVAL;
874 SYSCALL_DEFINE3(listxattr, const char __user *, pathname, char __user *, list,
877 return path_listxattr(pathname, list, size, LOOKUP_FOLLOW);
880 SYSCALL_DEFINE3(llistxattr, const char __user *, pathname, char __user *, list,
883 return path_listxattr(pathname, list, size, 0);
886 SYSCALL_DEFINE3(flistxattr, int, fd, char __user *, list, size_t, size)
888 struct fd f = fdget(fd);
889 ssize_t error = -EBADF;
894 error = listxattr(f.file->f_path.dentry, list, size);
900 * Extended attribute REMOVE operations
903 removexattr(struct mnt_idmap *idmap, struct dentry *d, const char *name)
905 if (is_posix_acl_xattr(name))
906 return vfs_remove_acl(idmap, d, name);
907 return vfs_removexattr(idmap, d, name);
910 static int path_removexattr(const char __user *pathname,
911 const char __user *name, unsigned int lookup_flags)
915 char kname[XATTR_NAME_MAX + 1];
917 error = strncpy_from_user(kname, name, sizeof(kname));
918 if (error == 0 || error == sizeof(kname))
923 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
926 error = mnt_want_write(path.mnt);
928 error = removexattr(mnt_idmap(path.mnt), path.dentry, kname);
929 mnt_drop_write(path.mnt);
932 if (retry_estale(error, lookup_flags)) {
933 lookup_flags |= LOOKUP_REVAL;
939 SYSCALL_DEFINE2(removexattr, const char __user *, pathname,
940 const char __user *, name)
942 return path_removexattr(pathname, name, LOOKUP_FOLLOW);
945 SYSCALL_DEFINE2(lremovexattr, const char __user *, pathname,
946 const char __user *, name)
948 return path_removexattr(pathname, name, 0);
951 SYSCALL_DEFINE2(fremovexattr, int, fd, const char __user *, name)
953 struct fd f = fdget(fd);
954 char kname[XATTR_NAME_MAX + 1];
961 error = strncpy_from_user(kname, name, sizeof(kname));
962 if (error == 0 || error == sizeof(kname))
967 error = mnt_want_write_file(f.file);
969 error = removexattr(file_mnt_idmap(f.file),
970 f.file->f_path.dentry, kname);
971 mnt_drop_write_file(f.file);
977 int xattr_list_one(char **buffer, ssize_t *remaining_size, const char *name)
981 len = strlen(name) + 1;
983 if (*remaining_size < len)
985 memcpy(*buffer, name, len);
988 *remaining_size -= len;
993 * generic_listxattr - run through a dentry's xattr list() operations
994 * @dentry: dentry to list the xattrs
995 * @buffer: result buffer
996 * @buffer_size: size of @buffer
998 * Combine the results of the list() operation from every xattr_handler in the
999 * xattr_handler stack.
1001 * Note that this will not include the entries for POSIX ACLs.
1004 generic_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size)
1006 const struct xattr_handler *handler, * const *handlers = dentry->d_sb->s_xattr;
1007 ssize_t remaining_size = buffer_size;
1010 for_each_xattr_handler(handlers, handler) {
1011 if (!handler->name || (handler->list && !handler->list(dentry)))
1013 err = xattr_list_one(&buffer, &remaining_size, handler->name);
1018 return err ? err : buffer_size - remaining_size;
1020 EXPORT_SYMBOL(generic_listxattr);
1023 * xattr_full_name - Compute full attribute name from suffix
1025 * @handler: handler of the xattr_handler operation
1026 * @name: name passed to the xattr_handler operation
1028 * The get and set xattr handler operations are called with the remainder of
1029 * the attribute name after skipping the handler's prefix: for example, "foo"
1030 * is passed to the get operation of a handler with prefix "user." to get
1031 * attribute "user.foo". The full name is still "there" in the name though.
1033 * Note: the list xattr handler operation when called from the vfs is passed a
1034 * NULL name; some file systems use this operation internally, with varying
1037 const char *xattr_full_name(const struct xattr_handler *handler,
1040 size_t prefix_len = strlen(xattr_prefix(handler));
1042 return name - prefix_len;
1044 EXPORT_SYMBOL(xattr_full_name);
1047 * simple_xattr_space - estimate the memory used by a simple xattr
1048 * @name: the full name of the xattr
1049 * @size: the size of its value
1051 * This takes no account of how much larger the two slab objects actually are:
1052 * that would depend on the slab implementation, when what is required is a
1053 * deterministic number, which grows with name length and size and quantity.
1055 * Return: The approximate number of bytes of memory used by such an xattr.
1057 size_t simple_xattr_space(const char *name, size_t size)
1060 * Use "40" instead of sizeof(struct simple_xattr), to return the
1061 * same result on 32-bit and 64-bit, and even if simple_xattr grows.
1063 return 40 + size + strlen(name);
1067 * simple_xattr_free - free an xattr object
1068 * @xattr: the xattr object
1070 * Free the xattr object. Can handle @xattr being NULL.
1072 void simple_xattr_free(struct simple_xattr *xattr)
1080 * simple_xattr_alloc - allocate new xattr object
1081 * @value: value of the xattr object
1082 * @size: size of @value
1084 * Allocate a new xattr object and initialize respective members. The caller is
1085 * responsible for handling the name of the xattr.
1087 * Return: On success a new xattr object is returned. On failure NULL is
1090 struct simple_xattr *simple_xattr_alloc(const void *value, size_t size)
1092 struct simple_xattr *new_xattr;
1096 len = sizeof(*new_xattr) + size;
1097 if (len < sizeof(*new_xattr))
1100 new_xattr = kvmalloc(len, GFP_KERNEL_ACCOUNT);
1104 new_xattr->size = size;
1105 memcpy(new_xattr->value, value, size);
1110 * rbtree_simple_xattr_cmp - compare xattr name with current rbtree xattr entry
1112 * @node: current node
1114 * Compare the xattr name with the xattr name attached to @node in the rbtree.
1116 * Return: Negative value if continuing left, positive if continuing right, 0
1117 * if the xattr attached to @node matches @key.
1119 static int rbtree_simple_xattr_cmp(const void *key, const struct rb_node *node)
1121 const char *xattr_name = key;
1122 const struct simple_xattr *xattr;
1124 xattr = rb_entry(node, struct simple_xattr, rb_node);
1125 return strcmp(xattr->name, xattr_name);
1129 * rbtree_simple_xattr_node_cmp - compare two xattr rbtree nodes
1130 * @new_node: new node
1131 * @node: current node
1133 * Compare the xattr attached to @new_node with the xattr attached to @node.
1135 * Return: Negative value if continuing left, positive if continuing right, 0
1136 * if the xattr attached to @new_node matches the xattr attached to @node.
1138 static int rbtree_simple_xattr_node_cmp(struct rb_node *new_node,
1139 const struct rb_node *node)
1141 struct simple_xattr *xattr;
1142 xattr = rb_entry(new_node, struct simple_xattr, rb_node);
1143 return rbtree_simple_xattr_cmp(xattr->name, node);
1147 * simple_xattr_get - get an xattr object
1148 * @xattrs: the header of the xattr object
1149 * @name: the name of the xattr to retrieve
1150 * @buffer: the buffer to store the value into
1151 * @size: the size of @buffer
1153 * Try to find and retrieve the xattr object associated with @name.
1154 * If @buffer is provided store the value of @xattr in @buffer
1155 * otherwise just return the length. The size of @buffer is limited
1156 * to XATTR_SIZE_MAX which currently is 65536.
1158 * Return: On success the length of the xattr value is returned. On error a
1159 * negative error code is returned.
1161 int simple_xattr_get(struct simple_xattrs *xattrs, const char *name,
1162 void *buffer, size_t size)
1164 struct simple_xattr *xattr = NULL;
1165 struct rb_node *rbp;
1168 read_lock(&xattrs->lock);
1169 rbp = rb_find(name, &xattrs->rb_root, rbtree_simple_xattr_cmp);
1171 xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1174 if (size < xattr->size)
1177 memcpy(buffer, xattr->value, xattr->size);
1180 read_unlock(&xattrs->lock);
1185 * simple_xattr_set - set an xattr object
1186 * @xattrs: the header of the xattr object
1187 * @name: the name of the xattr to retrieve
1188 * @value: the value to store along the xattr
1189 * @size: the size of @value
1190 * @flags: the flags determining how to set the xattr
1192 * Set a new xattr object.
1193 * If @value is passed a new xattr object will be allocated. If XATTR_REPLACE
1194 * is specified in @flags a matching xattr object for @name must already exist.
1195 * If it does it will be replaced with the new xattr object. If it doesn't we
1196 * fail. If XATTR_CREATE is specified and a matching xattr does already exist
1197 * we fail. If it doesn't we create a new xattr. If @flags is zero we simply
1198 * insert the new xattr replacing any existing one.
1200 * If @value is empty and a matching xattr object is found we delete it if
1201 * XATTR_REPLACE is specified in @flags or @flags is zero.
1203 * If @value is empty and no matching xattr object for @name is found we do
1204 * nothing if XATTR_CREATE is specified in @flags or @flags is zero. For
1205 * XATTR_REPLACE we fail as mentioned above.
1207 * Return: On success, the removed or replaced xattr is returned, to be freed
1208 * by the caller; or NULL if none. On failure a negative error code is returned.
1210 struct simple_xattr *simple_xattr_set(struct simple_xattrs *xattrs,
1211 const char *name, const void *value,
1212 size_t size, int flags)
1214 struct simple_xattr *old_xattr = NULL, *new_xattr = NULL;
1215 struct rb_node *parent = NULL, **rbp;
1218 /* value == NULL means remove */
1220 new_xattr = simple_xattr_alloc(value, size);
1222 return ERR_PTR(-ENOMEM);
1224 new_xattr->name = kstrdup(name, GFP_KERNEL_ACCOUNT);
1225 if (!new_xattr->name) {
1226 simple_xattr_free(new_xattr);
1227 return ERR_PTR(-ENOMEM);
1231 write_lock(&xattrs->lock);
1232 rbp = &xattrs->rb_root.rb_node;
1235 ret = rbtree_simple_xattr_cmp(name, *rbp);
1237 rbp = &(*rbp)->rb_left;
1239 rbp = &(*rbp)->rb_right;
1241 old_xattr = rb_entry(*rbp, struct simple_xattr, rb_node);
1247 /* Fail if XATTR_CREATE is requested and the xattr exists. */
1248 if (flags & XATTR_CREATE) {
1254 rb_replace_node(&old_xattr->rb_node,
1255 &new_xattr->rb_node, &xattrs->rb_root);
1257 rb_erase(&old_xattr->rb_node, &xattrs->rb_root);
1259 /* Fail if XATTR_REPLACE is requested but no xattr is found. */
1260 if (flags & XATTR_REPLACE) {
1266 * If XATTR_CREATE or no flags are specified together with a
1267 * new value simply insert it.
1270 rb_link_node(&new_xattr->rb_node, parent, rbp);
1271 rb_insert_color(&new_xattr->rb_node, &xattrs->rb_root);
1275 * If XATTR_CREATE or no flags are specified and neither an
1276 * old or new xattr exist then we don't need to do anything.
1281 write_unlock(&xattrs->lock);
1284 simple_xattr_free(new_xattr);
1285 return ERR_PTR(err);
1288 static bool xattr_is_trusted(const char *name)
1290 return !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN);
1294 * simple_xattr_list - list all xattr objects
1295 * @inode: inode from which to get the xattrs
1296 * @xattrs: the header of the xattr object
1297 * @buffer: the buffer to store all xattrs into
1298 * @size: the size of @buffer
1300 * List all xattrs associated with @inode. If @buffer is NULL we returned
1301 * the required size of the buffer. If @buffer is provided we store the
1302 * xattrs value into it provided it is big enough.
1304 * Note, the number of xattr names that can be listed with listxattr(2) is
1305 * limited to XATTR_LIST_MAX aka 65536 bytes. If a larger buffer is passed
1306 * then vfs_listxattr() caps it to XATTR_LIST_MAX and if more xattr names
1307 * are found it will return -E2BIG.
1309 * Return: On success the required size or the size of the copied xattrs is
1310 * returned. On error a negative error code is returned.
1312 ssize_t simple_xattr_list(struct inode *inode, struct simple_xattrs *xattrs,
1313 char *buffer, size_t size)
1315 bool trusted = ns_capable_noaudit(&init_user_ns, CAP_SYS_ADMIN);
1316 struct simple_xattr *xattr;
1317 struct rb_node *rbp;
1318 ssize_t remaining_size = size;
1321 err = posix_acl_listxattr(inode, &buffer, &remaining_size);
1325 read_lock(&xattrs->lock);
1326 for (rbp = rb_first(&xattrs->rb_root); rbp; rbp = rb_next(rbp)) {
1327 xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1329 /* skip "trusted." attributes for unprivileged callers */
1330 if (!trusted && xattr_is_trusted(xattr->name))
1333 err = xattr_list_one(&buffer, &remaining_size, xattr->name);
1337 read_unlock(&xattrs->lock);
1339 return err ? err : size - remaining_size;
1343 * rbtree_simple_xattr_less - compare two xattr rbtree nodes
1344 * @new_node: new node
1345 * @node: current node
1347 * Compare the xattr attached to @new_node with the xattr attached to @node.
1348 * Note that this function technically tolerates duplicate entries.
1350 * Return: True if insertion point in the rbtree is found.
1352 static bool rbtree_simple_xattr_less(struct rb_node *new_node,
1353 const struct rb_node *node)
1355 return rbtree_simple_xattr_node_cmp(new_node, node) < 0;
1359 * simple_xattr_add - add xattr objects
1360 * @xattrs: the header of the xattr object
1361 * @new_xattr: the xattr object to add
1363 * Add an xattr object to @xattrs. This assumes no replacement or removal
1364 * of matching xattrs is wanted. Should only be called during inode
1365 * initialization when a few distinct initial xattrs are supposed to be set.
1367 void simple_xattr_add(struct simple_xattrs *xattrs,
1368 struct simple_xattr *new_xattr)
1370 write_lock(&xattrs->lock);
1371 rb_add(&new_xattr->rb_node, &xattrs->rb_root, rbtree_simple_xattr_less);
1372 write_unlock(&xattrs->lock);
1376 * simple_xattrs_init - initialize new xattr header
1377 * @xattrs: header to initialize
1379 * Initialize relevant fields of a an xattr header.
1381 void simple_xattrs_init(struct simple_xattrs *xattrs)
1383 xattrs->rb_root = RB_ROOT;
1384 rwlock_init(&xattrs->lock);
1388 * simple_xattrs_free - free xattrs
1389 * @xattrs: xattr header whose xattrs to destroy
1390 * @freed_space: approximate number of bytes of memory freed from @xattrs
1392 * Destroy all xattrs in @xattr. When this is called no one can hold a
1393 * reference to any of the xattrs anymore.
1395 void simple_xattrs_free(struct simple_xattrs *xattrs, size_t *freed_space)
1397 struct rb_node *rbp;
1401 rbp = rb_first(&xattrs->rb_root);
1403 struct simple_xattr *xattr;
1404 struct rb_node *rbp_next;
1406 rbp_next = rb_next(rbp);
1407 xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1408 rb_erase(&xattr->rb_node, &xattrs->rb_root);
1410 *freed_space += simple_xattr_space(xattr->name,
1412 simple_xattr_free(xattr);