1 // SPDX-License-Identifier: GPL-2.0-only
5 Extended attribute handling.
7 Copyright (C) 2001 by Andreas Gruenbacher <a.gruenbacher@computer.org>
8 Copyright (C) 2001 SGI - Silicon Graphics, Inc <linux-xfs@oss.sgi.com>
9 Copyright (c) 2004 Red Hat, Inc., James Morris <jmorris@redhat.com>
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/evm.h>
20 #include <linux/syscalls.h>
21 #include <linux/export.h>
22 #include <linux/fsnotify.h>
23 #include <linux/audit.h>
24 #include <linux/vmalloc.h>
25 #include <linux/posix_acl_xattr.h>
27 #include <linux/uaccess.h>
32 strcmp_prefix(const char *a, const char *a_prefix)
34 while (*a_prefix && *a == *a_prefix) {
38 return *a_prefix ? NULL : a;
42 * In order to implement different sets of xattr operations for each xattr
43 * prefix, a filesystem should create a null-terminated array of struct
44 * xattr_handler (one for each prefix) and hang a pointer to it off of the
45 * s_xattr field of the superblock.
47 #define for_each_xattr_handler(handlers, handler) \
49 for ((handler) = *(handlers)++; \
51 (handler) = *(handlers)++)
54 * Find the xattr_handler with the matching prefix.
56 static const struct xattr_handler *
57 xattr_resolve_name(struct inode *inode, const char **name)
59 const struct xattr_handler * const *handlers = inode->i_sb->s_xattr;
60 const struct xattr_handler *handler;
62 if (!(inode->i_opflags & IOP_XATTR)) {
63 if (unlikely(is_bad_inode(inode)))
65 return ERR_PTR(-EOPNOTSUPP);
67 for_each_xattr_handler(handlers, handler) {
70 n = strcmp_prefix(*name, xattr_prefix(handler));
72 if (!handler->prefix ^ !*n) {
75 return ERR_PTR(-EINVAL);
81 return ERR_PTR(-EOPNOTSUPP);
85 * may_write_xattr - check whether inode allows writing xattr
86 * @idmap: idmap of the mount the inode was found from
87 * @inode: the inode on which to set an xattr
89 * Check whether the inode allows writing xattrs. Specifically, we can never
90 * set or remove an extended attribute on a read-only filesystem or on an
91 * immutable / append-only inode.
93 * We also need to ensure that the inode has a mapping in the mount to
94 * not risk writing back invalid i_{g,u}id values.
96 * Return: On success zero is returned. On error a negative errno is returned.
98 int may_write_xattr(struct mnt_idmap *idmap, struct inode *inode)
100 if (IS_IMMUTABLE(inode))
102 if (IS_APPEND(inode))
104 if (HAS_UNMAPPED_ID(idmap, inode))
110 * Check permissions for extended attribute access. This is a bit complicated
111 * because different namespaces have very different rules.
114 xattr_permission(struct mnt_idmap *idmap, struct inode *inode,
115 const char *name, int mask)
117 if (mask & MAY_WRITE) {
120 ret = may_write_xattr(idmap, inode);
126 * No restriction for security.* and system.* from the VFS. Decision
127 * on these is left to the underlying filesystem / security module.
129 if (!strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN) ||
130 !strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
134 * The trusted.* namespace can only be accessed by privileged users.
136 if (!strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN)) {
137 if (!capable(CAP_SYS_ADMIN))
138 return (mask & MAY_WRITE) ? -EPERM : -ENODATA;
143 * In the user.* namespace, only regular files and directories can have
144 * extended attributes. For sticky directories, only the owner and
145 * privileged users can write attributes.
147 if (!strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN)) {
148 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
149 return (mask & MAY_WRITE) ? -EPERM : -ENODATA;
150 if (S_ISDIR(inode->i_mode) && (inode->i_mode & S_ISVTX) &&
151 (mask & MAY_WRITE) &&
152 !inode_owner_or_capable(idmap, inode))
156 return inode_permission(idmap, inode, mask);
160 * Look for any handler that deals with the specified namespace.
163 xattr_supports_user_prefix(struct inode *inode)
165 const struct xattr_handler * const *handlers = inode->i_sb->s_xattr;
166 const struct xattr_handler *handler;
168 if (!(inode->i_opflags & IOP_XATTR)) {
169 if (unlikely(is_bad_inode(inode)))
174 for_each_xattr_handler(handlers, handler) {
175 if (!strncmp(xattr_prefix(handler), XATTR_USER_PREFIX,
176 XATTR_USER_PREFIX_LEN))
182 EXPORT_SYMBOL(xattr_supports_user_prefix);
185 __vfs_setxattr(struct mnt_idmap *idmap, struct dentry *dentry,
186 struct inode *inode, const char *name, const void *value,
187 size_t size, int flags)
189 const struct xattr_handler *handler;
191 if (is_posix_acl_xattr(name))
194 handler = xattr_resolve_name(inode, &name);
196 return PTR_ERR(handler);
200 value = ""; /* empty EA, do not remove */
201 return handler->set(handler, idmap, dentry, inode, name, value,
204 EXPORT_SYMBOL(__vfs_setxattr);
207 * __vfs_setxattr_noperm - perform setxattr operation without performing
210 * @idmap: idmap of the mount the inode was found from
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
217 * returns the result of the internal setxattr or setsecurity operations.
219 * This function requires the caller to lock the inode's i_mutex before it
220 * is executed. It also assumes that the caller will make the appropriate
223 int __vfs_setxattr_noperm(struct mnt_idmap *idmap,
224 struct dentry *dentry, const char *name,
225 const void *value, size_t size, int flags)
227 struct inode *inode = dentry->d_inode;
229 int issec = !strncmp(name, XATTR_SECURITY_PREFIX,
230 XATTR_SECURITY_PREFIX_LEN);
233 inode->i_flags &= ~S_NOSEC;
234 if (inode->i_opflags & IOP_XATTR) {
235 error = __vfs_setxattr(idmap, dentry, inode, name, value,
238 fsnotify_xattr(dentry);
239 security_inode_post_setxattr(dentry, name, value,
243 if (unlikely(is_bad_inode(inode)))
246 if (error == -EAGAIN) {
250 const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
252 error = security_inode_setsecurity(inode, suffix, value,
255 fsnotify_xattr(dentry);
263 * __vfs_setxattr_locked - set an extended attribute while holding the inode
266 * @idmap: idmap of the mount of the target inode
267 * @dentry: object to perform setxattr on
268 * @name: xattr name to set
269 * @value: value to set @name to
270 * @size: size of @value
271 * @flags: flags to pass into filesystem operations
272 * @delegated_inode: on return, will contain an inode pointer that
273 * a delegation was broken on, NULL if none.
276 __vfs_setxattr_locked(struct mnt_idmap *idmap, struct dentry *dentry,
277 const char *name, const void *value, size_t size,
278 int flags, struct inode **delegated_inode)
280 struct inode *inode = dentry->d_inode;
283 error = xattr_permission(idmap, inode, name, MAY_WRITE);
287 error = security_inode_setxattr(idmap, dentry, name, value, size,
292 error = try_break_deleg(inode, delegated_inode);
296 error = __vfs_setxattr_noperm(idmap, dentry, name, value,
302 EXPORT_SYMBOL_GPL(__vfs_setxattr_locked);
305 vfs_setxattr(struct mnt_idmap *idmap, struct dentry *dentry,
306 const char *name, const void *value, size_t size, int flags)
308 struct inode *inode = dentry->d_inode;
309 struct inode *delegated_inode = NULL;
310 const void *orig_value = value;
313 if (size && strcmp(name, XATTR_NAME_CAPS) == 0) {
314 error = cap_convert_nscap(idmap, dentry, &value, size);
322 error = __vfs_setxattr_locked(idmap, dentry, name, value, size,
323 flags, &delegated_inode);
326 if (delegated_inode) {
327 error = break_deleg_wait(&delegated_inode);
331 if (value != orig_value)
336 EXPORT_SYMBOL_GPL(vfs_setxattr);
339 xattr_getsecurity(struct mnt_idmap *idmap, struct inode *inode,
340 const char *name, void *value, size_t size)
345 if (!value || !size) {
346 len = security_inode_getsecurity(idmap, inode, name,
351 len = security_inode_getsecurity(idmap, inode, name, &buffer,
359 memcpy(value, buffer, len);
367 * vfs_getxattr_alloc - allocate memory, if necessary, before calling getxattr
369 * Allocate memory, if not already allocated, or re-allocate correct size,
370 * before retrieving the extended attribute. The xattr value buffer should
371 * always be freed by the caller, even on error.
373 * Returns the result of alloc, if failed, or the getxattr operation.
376 vfs_getxattr_alloc(struct mnt_idmap *idmap, struct dentry *dentry,
377 const char *name, char **xattr_value, size_t xattr_size,
380 const struct xattr_handler *handler;
381 struct inode *inode = dentry->d_inode;
382 char *value = *xattr_value;
385 error = xattr_permission(idmap, inode, name, MAY_READ);
389 handler = xattr_resolve_name(inode, &name);
391 return PTR_ERR(handler);
394 error = handler->get(handler, dentry, inode, name, NULL, 0);
398 if (!value || (error > xattr_size)) {
399 value = krealloc(*xattr_value, error + 1, flags);
402 memset(value, 0, error + 1);
405 error = handler->get(handler, dentry, inode, name, value, error);
406 *xattr_value = value;
411 __vfs_getxattr(struct dentry *dentry, struct inode *inode, const char *name,
412 void *value, size_t size)
414 const struct xattr_handler *handler;
416 if (is_posix_acl_xattr(name))
419 handler = xattr_resolve_name(inode, &name);
421 return PTR_ERR(handler);
424 return handler->get(handler, dentry, inode, name, value, size);
426 EXPORT_SYMBOL(__vfs_getxattr);
429 vfs_getxattr(struct mnt_idmap *idmap, struct dentry *dentry,
430 const char *name, void *value, size_t size)
432 struct inode *inode = dentry->d_inode;
435 error = xattr_permission(idmap, inode, name, MAY_READ);
439 error = security_inode_getxattr(dentry, name);
443 if (!strncmp(name, XATTR_SECURITY_PREFIX,
444 XATTR_SECURITY_PREFIX_LEN)) {
445 const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
446 int ret = xattr_getsecurity(idmap, inode, suffix, value,
449 * Only overwrite the return value if a security module
450 * is actually active.
452 if (ret == -EOPNOTSUPP)
457 return __vfs_getxattr(dentry, inode, name, value, size);
459 EXPORT_SYMBOL_GPL(vfs_getxattr);
462 * vfs_listxattr - retrieve \0 separated list of xattr names
463 * @dentry: the dentry from whose inode the xattr names are retrieved
464 * @list: buffer to store xattr names into
465 * @size: size of the buffer
467 * This function returns the names of all xattrs associated with the
470 * Note, for legacy reasons the vfs_listxattr() function lists POSIX
471 * ACLs as well. Since POSIX ACLs are decoupled from IOP_XATTR the
472 * vfs_listxattr() function doesn't check for this flag since a
473 * filesystem could implement POSIX ACLs without implementing any other
476 * However, since all codepaths that remove IOP_XATTR also assign of
477 * inode operations that either don't implement or implement a stub
478 * ->listxattr() operation.
480 * Return: On success, the size of the buffer that was used. On error a
481 * negative error code.
484 vfs_listxattr(struct dentry *dentry, char *list, size_t size)
486 struct inode *inode = d_inode(dentry);
489 error = security_inode_listxattr(dentry);
493 if (inode->i_op->listxattr) {
494 error = inode->i_op->listxattr(dentry, list, size);
496 error = security_inode_listsecurity(inode, list, size);
497 if (size && error > size)
502 EXPORT_SYMBOL_GPL(vfs_listxattr);
505 __vfs_removexattr(struct mnt_idmap *idmap, struct dentry *dentry,
508 struct inode *inode = d_inode(dentry);
509 const struct xattr_handler *handler;
511 if (is_posix_acl_xattr(name))
514 handler = xattr_resolve_name(inode, &name);
516 return PTR_ERR(handler);
519 return handler->set(handler, idmap, dentry, inode, name, NULL, 0,
522 EXPORT_SYMBOL(__vfs_removexattr);
525 * __vfs_removexattr_locked - set an extended attribute while holding the inode
528 * @idmap: idmap of the mount of the target inode
529 * @dentry: object to perform setxattr on
530 * @name: name of xattr to remove
531 * @delegated_inode: on return, will contain an inode pointer that
532 * a delegation was broken on, NULL if none.
535 __vfs_removexattr_locked(struct mnt_idmap *idmap,
536 struct dentry *dentry, const char *name,
537 struct inode **delegated_inode)
539 struct inode *inode = dentry->d_inode;
542 error = xattr_permission(idmap, inode, name, MAY_WRITE);
546 error = security_inode_removexattr(idmap, dentry, name);
550 error = try_break_deleg(inode, delegated_inode);
554 error = __vfs_removexattr(idmap, dentry, name);
557 fsnotify_xattr(dentry);
558 evm_inode_post_removexattr(dentry, name);
564 EXPORT_SYMBOL_GPL(__vfs_removexattr_locked);
567 vfs_removexattr(struct mnt_idmap *idmap, struct dentry *dentry,
570 struct inode *inode = dentry->d_inode;
571 struct inode *delegated_inode = NULL;
576 error = __vfs_removexattr_locked(idmap, dentry,
577 name, &delegated_inode);
580 if (delegated_inode) {
581 error = break_deleg_wait(&delegated_inode);
588 EXPORT_SYMBOL_GPL(vfs_removexattr);
591 * Extended attribute SET operations
594 int setxattr_copy(const char __user *name, struct xattr_ctx *ctx)
598 if (ctx->flags & ~(XATTR_CREATE|XATTR_REPLACE))
601 error = strncpy_from_user(ctx->kname->name, name,
602 sizeof(ctx->kname->name));
603 if (error == 0 || error == sizeof(ctx->kname->name))
610 if (ctx->size > XATTR_SIZE_MAX)
613 ctx->kvalue = vmemdup_user(ctx->cvalue, ctx->size);
614 if (IS_ERR(ctx->kvalue)) {
615 error = PTR_ERR(ctx->kvalue);
623 int do_setxattr(struct mnt_idmap *idmap, struct dentry *dentry,
624 struct xattr_ctx *ctx)
626 if (is_posix_acl_xattr(ctx->kname->name))
627 return do_set_acl(idmap, dentry, ctx->kname->name,
628 ctx->kvalue, ctx->size);
630 return vfs_setxattr(idmap, dentry, ctx->kname->name,
631 ctx->kvalue, ctx->size, ctx->flags);
635 setxattr(struct mnt_idmap *idmap, struct dentry *d,
636 const char __user *name, const void __user *value, size_t size,
639 struct xattr_name kname;
640 struct xattr_ctx ctx = {
649 error = setxattr_copy(name, &ctx);
653 error = do_setxattr(idmap, d, &ctx);
659 static int path_setxattr(const char __user *pathname,
660 const char __user *name, const void __user *value,
661 size_t size, int flags, unsigned int lookup_flags)
667 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
670 error = mnt_want_write(path.mnt);
672 error = setxattr(mnt_idmap(path.mnt), path.dentry, name,
674 mnt_drop_write(path.mnt);
677 if (retry_estale(error, lookup_flags)) {
678 lookup_flags |= LOOKUP_REVAL;
684 SYSCALL_DEFINE5(setxattr, const char __user *, pathname,
685 const char __user *, name, const void __user *, value,
686 size_t, size, int, flags)
688 return path_setxattr(pathname, name, value, size, flags, LOOKUP_FOLLOW);
691 SYSCALL_DEFINE5(lsetxattr, const char __user *, pathname,
692 const char __user *, name, const void __user *, value,
693 size_t, size, int, flags)
695 return path_setxattr(pathname, name, value, size, flags, 0);
698 SYSCALL_DEFINE5(fsetxattr, int, fd, const char __user *, name,
699 const void __user *,value, size_t, size, int, flags)
701 struct fd f = fdget(fd);
707 error = mnt_want_write_file(f.file);
709 error = setxattr(file_mnt_idmap(f.file),
710 f.file->f_path.dentry, name,
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,
904 const char __user *name)
907 char kname[XATTR_NAME_MAX + 1];
909 error = strncpy_from_user(kname, name, sizeof(kname));
910 if (error == 0 || error == sizeof(kname))
915 if (is_posix_acl_xattr(kname))
916 return vfs_remove_acl(idmap, d, kname);
918 return vfs_removexattr(idmap, d, kname);
921 static int path_removexattr(const char __user *pathname,
922 const char __user *name, unsigned int lookup_flags)
927 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
930 error = mnt_want_write(path.mnt);
932 error = removexattr(mnt_idmap(path.mnt), path.dentry, name);
933 mnt_drop_write(path.mnt);
936 if (retry_estale(error, lookup_flags)) {
937 lookup_flags |= LOOKUP_REVAL;
943 SYSCALL_DEFINE2(removexattr, const char __user *, pathname,
944 const char __user *, name)
946 return path_removexattr(pathname, name, LOOKUP_FOLLOW);
949 SYSCALL_DEFINE2(lremovexattr, const char __user *, pathname,
950 const char __user *, name)
952 return path_removexattr(pathname, name, 0);
955 SYSCALL_DEFINE2(fremovexattr, int, fd, const char __user *, name)
957 struct fd f = fdget(fd);
963 error = mnt_want_write_file(f.file);
965 error = removexattr(file_mnt_idmap(f.file),
966 f.file->f_path.dentry, name);
967 mnt_drop_write_file(f.file);
973 int xattr_list_one(char **buffer, ssize_t *remaining_size, const char *name)
977 len = strlen(name) + 1;
979 if (*remaining_size < len)
981 memcpy(*buffer, name, len);
984 *remaining_size -= len;
989 * generic_listxattr - run through a dentry's xattr list() operations
990 * @dentry: dentry to list the xattrs
991 * @buffer: result buffer
992 * @buffer_size: size of @buffer
994 * Combine the results of the list() operation from every xattr_handler in the
995 * xattr_handler stack.
997 * Note that this will not include the entries for POSIX ACLs.
1000 generic_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size)
1002 const struct xattr_handler *handler, * const *handlers = dentry->d_sb->s_xattr;
1003 ssize_t remaining_size = buffer_size;
1006 for_each_xattr_handler(handlers, handler) {
1007 if (!handler->name || (handler->list && !handler->list(dentry)))
1009 err = xattr_list_one(&buffer, &remaining_size, handler->name);
1014 return err ? err : buffer_size - remaining_size;
1016 EXPORT_SYMBOL(generic_listxattr);
1019 * xattr_full_name - Compute full attribute name from suffix
1021 * @handler: handler of the xattr_handler operation
1022 * @name: name passed to the xattr_handler operation
1024 * The get and set xattr handler operations are called with the remainder of
1025 * the attribute name after skipping the handler's prefix: for example, "foo"
1026 * is passed to the get operation of a handler with prefix "user." to get
1027 * attribute "user.foo". The full name is still "there" in the name though.
1029 * Note: the list xattr handler operation when called from the vfs is passed a
1030 * NULL name; some file systems use this operation internally, with varying
1033 const char *xattr_full_name(const struct xattr_handler *handler,
1036 size_t prefix_len = strlen(xattr_prefix(handler));
1038 return name - prefix_len;
1040 EXPORT_SYMBOL(xattr_full_name);
1043 * simple_xattr_space - estimate the memory used by a simple xattr
1044 * @name: the full name of the xattr
1045 * @size: the size of its value
1047 * This takes no account of how much larger the two slab objects actually are:
1048 * that would depend on the slab implementation, when what is required is a
1049 * deterministic number, which grows with name length and size and quantity.
1051 * Return: The approximate number of bytes of memory used by such an xattr.
1053 size_t simple_xattr_space(const char *name, size_t size)
1056 * Use "40" instead of sizeof(struct simple_xattr), to return the
1057 * same result on 32-bit and 64-bit, and even if simple_xattr grows.
1059 return 40 + size + strlen(name);
1063 * simple_xattr_free - free an xattr object
1064 * @xattr: the xattr object
1066 * Free the xattr object. Can handle @xattr being NULL.
1068 void simple_xattr_free(struct simple_xattr *xattr)
1076 * simple_xattr_alloc - allocate new xattr object
1077 * @value: value of the xattr object
1078 * @size: size of @value
1080 * Allocate a new xattr object and initialize respective members. The caller is
1081 * responsible for handling the name of the xattr.
1083 * Return: On success a new xattr object is returned. On failure NULL is
1086 struct simple_xattr *simple_xattr_alloc(const void *value, size_t size)
1088 struct simple_xattr *new_xattr;
1092 len = sizeof(*new_xattr) + size;
1093 if (len < sizeof(*new_xattr))
1096 new_xattr = kvmalloc(len, GFP_KERNEL_ACCOUNT);
1100 new_xattr->size = size;
1101 memcpy(new_xattr->value, value, size);
1106 * rbtree_simple_xattr_cmp - compare xattr name with current rbtree xattr entry
1108 * @node: current node
1110 * Compare the xattr name with the xattr name attached to @node in the rbtree.
1112 * Return: Negative value if continuing left, positive if continuing right, 0
1113 * if the xattr attached to @node matches @key.
1115 static int rbtree_simple_xattr_cmp(const void *key, const struct rb_node *node)
1117 const char *xattr_name = key;
1118 const struct simple_xattr *xattr;
1120 xattr = rb_entry(node, struct simple_xattr, rb_node);
1121 return strcmp(xattr->name, xattr_name);
1125 * rbtree_simple_xattr_node_cmp - compare two xattr rbtree nodes
1126 * @new_node: new node
1127 * @node: current node
1129 * Compare the xattr attached to @new_node with the xattr attached to @node.
1131 * Return: Negative value if continuing left, positive if continuing right, 0
1132 * if the xattr attached to @new_node matches the xattr attached to @node.
1134 static int rbtree_simple_xattr_node_cmp(struct rb_node *new_node,
1135 const struct rb_node *node)
1137 struct simple_xattr *xattr;
1138 xattr = rb_entry(new_node, struct simple_xattr, rb_node);
1139 return rbtree_simple_xattr_cmp(xattr->name, node);
1143 * simple_xattr_get - get an xattr object
1144 * @xattrs: the header of the xattr object
1145 * @name: the name of the xattr to retrieve
1146 * @buffer: the buffer to store the value into
1147 * @size: the size of @buffer
1149 * Try to find and retrieve the xattr object associated with @name.
1150 * If @buffer is provided store the value of @xattr in @buffer
1151 * otherwise just return the length. The size of @buffer is limited
1152 * to XATTR_SIZE_MAX which currently is 65536.
1154 * Return: On success the length of the xattr value is returned. On error a
1155 * negative error code is returned.
1157 int simple_xattr_get(struct simple_xattrs *xattrs, const char *name,
1158 void *buffer, size_t size)
1160 struct simple_xattr *xattr = NULL;
1161 struct rb_node *rbp;
1164 read_lock(&xattrs->lock);
1165 rbp = rb_find(name, &xattrs->rb_root, rbtree_simple_xattr_cmp);
1167 xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1170 if (size < xattr->size)
1173 memcpy(buffer, xattr->value, xattr->size);
1176 read_unlock(&xattrs->lock);
1181 * simple_xattr_set - set an xattr object
1182 * @xattrs: the header of the xattr object
1183 * @name: the name of the xattr to retrieve
1184 * @value: the value to store along the xattr
1185 * @size: the size of @value
1186 * @flags: the flags determining how to set the xattr
1188 * Set a new xattr object.
1189 * If @value is passed a new xattr object will be allocated. If XATTR_REPLACE
1190 * is specified in @flags a matching xattr object for @name must already exist.
1191 * If it does it will be replaced with the new xattr object. If it doesn't we
1192 * fail. If XATTR_CREATE is specified and a matching xattr does already exist
1193 * we fail. If it doesn't we create a new xattr. If @flags is zero we simply
1194 * insert the new xattr replacing any existing one.
1196 * If @value is empty and a matching xattr object is found we delete it if
1197 * XATTR_REPLACE is specified in @flags or @flags is zero.
1199 * If @value is empty and no matching xattr object for @name is found we do
1200 * nothing if XATTR_CREATE is specified in @flags or @flags is zero. For
1201 * XATTR_REPLACE we fail as mentioned above.
1203 * Return: On success, the removed or replaced xattr is returned, to be freed
1204 * by the caller; or NULL if none. On failure a negative error code is returned.
1206 struct simple_xattr *simple_xattr_set(struct simple_xattrs *xattrs,
1207 const char *name, const void *value,
1208 size_t size, int flags)
1210 struct simple_xattr *old_xattr = NULL, *new_xattr = NULL;
1211 struct rb_node *parent = NULL, **rbp;
1214 /* value == NULL means remove */
1216 new_xattr = simple_xattr_alloc(value, size);
1218 return ERR_PTR(-ENOMEM);
1220 new_xattr->name = kstrdup(name, GFP_KERNEL_ACCOUNT);
1221 if (!new_xattr->name) {
1222 simple_xattr_free(new_xattr);
1223 return ERR_PTR(-ENOMEM);
1227 write_lock(&xattrs->lock);
1228 rbp = &xattrs->rb_root.rb_node;
1231 ret = rbtree_simple_xattr_cmp(name, *rbp);
1233 rbp = &(*rbp)->rb_left;
1235 rbp = &(*rbp)->rb_right;
1237 old_xattr = rb_entry(*rbp, struct simple_xattr, rb_node);
1243 /* Fail if XATTR_CREATE is requested and the xattr exists. */
1244 if (flags & XATTR_CREATE) {
1250 rb_replace_node(&old_xattr->rb_node,
1251 &new_xattr->rb_node, &xattrs->rb_root);
1253 rb_erase(&old_xattr->rb_node, &xattrs->rb_root);
1255 /* Fail if XATTR_REPLACE is requested but no xattr is found. */
1256 if (flags & XATTR_REPLACE) {
1262 * If XATTR_CREATE or no flags are specified together with a
1263 * new value simply insert it.
1266 rb_link_node(&new_xattr->rb_node, parent, rbp);
1267 rb_insert_color(&new_xattr->rb_node, &xattrs->rb_root);
1271 * If XATTR_CREATE or no flags are specified and neither an
1272 * old or new xattr exist then we don't need to do anything.
1277 write_unlock(&xattrs->lock);
1280 simple_xattr_free(new_xattr);
1281 return ERR_PTR(err);
1284 static bool xattr_is_trusted(const char *name)
1286 return !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN);
1290 * simple_xattr_list - list all xattr objects
1291 * @inode: inode from which to get the xattrs
1292 * @xattrs: the header of the xattr object
1293 * @buffer: the buffer to store all xattrs into
1294 * @size: the size of @buffer
1296 * List all xattrs associated with @inode. If @buffer is NULL we returned
1297 * the required size of the buffer. If @buffer is provided we store the
1298 * xattrs value into it provided it is big enough.
1300 * Note, the number of xattr names that can be listed with listxattr(2) is
1301 * limited to XATTR_LIST_MAX aka 65536 bytes. If a larger buffer is passed
1302 * then vfs_listxattr() caps it to XATTR_LIST_MAX and if more xattr names
1303 * are found it will return -E2BIG.
1305 * Return: On success the required size or the size of the copied xattrs is
1306 * returned. On error a negative error code is returned.
1308 ssize_t simple_xattr_list(struct inode *inode, struct simple_xattrs *xattrs,
1309 char *buffer, size_t size)
1311 bool trusted = ns_capable_noaudit(&init_user_ns, CAP_SYS_ADMIN);
1312 struct simple_xattr *xattr;
1313 struct rb_node *rbp;
1314 ssize_t remaining_size = size;
1317 err = posix_acl_listxattr(inode, &buffer, &remaining_size);
1321 read_lock(&xattrs->lock);
1322 for (rbp = rb_first(&xattrs->rb_root); rbp; rbp = rb_next(rbp)) {
1323 xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1325 /* skip "trusted." attributes for unprivileged callers */
1326 if (!trusted && xattr_is_trusted(xattr->name))
1329 err = xattr_list_one(&buffer, &remaining_size, xattr->name);
1333 read_unlock(&xattrs->lock);
1335 return err ? err : size - remaining_size;
1339 * rbtree_simple_xattr_less - compare two xattr rbtree nodes
1340 * @new_node: new node
1341 * @node: current node
1343 * Compare the xattr attached to @new_node with the xattr attached to @node.
1344 * Note that this function technically tolerates duplicate entries.
1346 * Return: True if insertion point in the rbtree is found.
1348 static bool rbtree_simple_xattr_less(struct rb_node *new_node,
1349 const struct rb_node *node)
1351 return rbtree_simple_xattr_node_cmp(new_node, node) < 0;
1355 * simple_xattr_add - add xattr objects
1356 * @xattrs: the header of the xattr object
1357 * @new_xattr: the xattr object to add
1359 * Add an xattr object to @xattrs. This assumes no replacement or removal
1360 * of matching xattrs is wanted. Should only be called during inode
1361 * initialization when a few distinct initial xattrs are supposed to be set.
1363 void simple_xattr_add(struct simple_xattrs *xattrs,
1364 struct simple_xattr *new_xattr)
1366 write_lock(&xattrs->lock);
1367 rb_add(&new_xattr->rb_node, &xattrs->rb_root, rbtree_simple_xattr_less);
1368 write_unlock(&xattrs->lock);
1372 * simple_xattrs_init - initialize new xattr header
1373 * @xattrs: header to initialize
1375 * Initialize relevant fields of a an xattr header.
1377 void simple_xattrs_init(struct simple_xattrs *xattrs)
1379 xattrs->rb_root = RB_ROOT;
1380 rwlock_init(&xattrs->lock);
1384 * simple_xattrs_free - free xattrs
1385 * @xattrs: xattr header whose xattrs to destroy
1386 * @freed_space: approximate number of bytes of memory freed from @xattrs
1388 * Destroy all xattrs in @xattr. When this is called no one can hold a
1389 * reference to any of the xattrs anymore.
1391 void simple_xattrs_free(struct simple_xattrs *xattrs, size_t *freed_space)
1393 struct rb_node *rbp;
1397 rbp = rb_first(&xattrs->rb_root);
1399 struct simple_xattr *xattr;
1400 struct rb_node *rbp_next;
1402 rbp_next = rb_next(rbp);
1403 xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1404 rb_erase(&xattr->rb_node, &xattrs->rb_root);
1406 *freed_space += simple_xattr_space(xattr->name,
1408 simple_xattr_free(xattr);