Merge tag 'timers_urgent_for_v6.13_rc1' of git://git.kernel.org/pub/scm/linux/kernel...
[drm/drm-misc.git] / fs / xattr.c
blob02bee149ad9674d3c16f6de137eb3c0fb7cabd89
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 File: fs/xattr.c
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>
11 #include <linux/fs.h>
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>
28 #include "internal.h"
30 static const char *
31 strcmp_prefix(const char *a, const char *a_prefix)
33 while (*a_prefix && *a == *a_prefix) {
34 a++;
35 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) \
47 if (handlers) \
48 for ((handler) = *(handlers)++; \
49 (handler) != NULL; \
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)))
63 return ERR_PTR(-EIO);
64 return ERR_PTR(-EOPNOTSUPP);
66 for_each_xattr_handler(handlers, handler) {
67 const char *n;
69 n = strcmp_prefix(*name, xattr_prefix(handler));
70 if (n) {
71 if (!handler->prefix ^ !*n) {
72 if (*n)
73 continue;
74 return ERR_PTR(-EINVAL);
76 *name = n;
77 return handler;
80 return ERR_PTR(-EOPNOTSUPP);
83 /**
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))
100 return -EPERM;
101 if (IS_APPEND(inode))
102 return -EPERM;
103 if (HAS_UNMAPPED_ID(idmap, inode))
104 return -EPERM;
105 return 0;
109 * Check permissions for extended attribute access. This is a bit complicated
110 * because different namespaces have very different rules.
112 static int
113 xattr_permission(struct mnt_idmap *idmap, struct inode *inode,
114 const char *name, int mask)
116 if (mask & MAY_WRITE) {
117 int ret;
119 ret = may_write_xattr(idmap, inode);
120 if (ret)
121 return ret;
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))
130 return 0;
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;
138 return 0;
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))
152 return -EPERM;
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)))
169 return -EIO;
170 return -EOPNOTSUPP;
173 for_each_xattr_handler(handlers, handler) {
174 if (!strncmp(xattr_prefix(handler), XATTR_USER_PREFIX,
175 XATTR_USER_PREFIX_LEN))
176 return 0;
179 return -EOPNOTSUPP;
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))
191 return -EOPNOTSUPP;
193 handler = xattr_resolve_name(inode, &name);
194 if (IS_ERR(handler))
195 return PTR_ERR(handler);
196 if (!handler->set)
197 return -EOPNOTSUPP;
198 if (size == 0)
199 value = ""; /* empty EA, do not remove */
200 return handler->set(handler, idmap, dentry, inode, name, value,
201 size, flags);
203 EXPORT_SYMBOL(__vfs_setxattr);
206 * __vfs_setxattr_noperm - perform setxattr operation without performing
207 * permission checks.
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
220 * permission checks.
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;
227 int error = -EAGAIN;
228 int issec = !strncmp(name, XATTR_SECURITY_PREFIX,
229 XATTR_SECURITY_PREFIX_LEN);
231 if (issec)
232 inode->i_flags &= ~S_NOSEC;
233 if (inode->i_opflags & IOP_XATTR) {
234 error = __vfs_setxattr(idmap, dentry, inode, name, value,
235 size, flags);
236 if (!error) {
237 fsnotify_xattr(dentry);
238 security_inode_post_setxattr(dentry, name, value,
239 size, flags);
241 } else {
242 if (unlikely(is_bad_inode(inode)))
243 return -EIO;
245 if (error == -EAGAIN) {
246 error = -EOPNOTSUPP;
248 if (issec) {
249 const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
251 error = security_inode_setsecurity(inode, suffix, value,
252 size, flags);
253 if (!error)
254 fsnotify_xattr(dentry);
258 return error;
262 * __vfs_setxattr_locked - set an extended attribute while holding the inode
263 * lock
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;
280 int error;
282 error = xattr_permission(idmap, inode, name, MAY_WRITE);
283 if (error)
284 return error;
286 error = security_inode_setxattr(idmap, dentry, name, value, size,
287 flags);
288 if (error)
289 goto out;
291 error = try_break_deleg(inode, delegated_inode);
292 if (error)
293 goto out;
295 error = __vfs_setxattr_noperm(idmap, dentry, name, value,
296 size, flags);
298 out:
299 return error;
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;
310 int error;
312 if (size && strcmp(name, XATTR_NAME_CAPS) == 0) {
313 error = cap_convert_nscap(idmap, dentry, &value, size);
314 if (error < 0)
315 return error;
316 size = error;
319 retry_deleg:
320 inode_lock(inode);
321 error = __vfs_setxattr_locked(idmap, dentry, name, value, size,
322 flags, &delegated_inode);
323 inode_unlock(inode);
325 if (delegated_inode) {
326 error = break_deleg_wait(&delegated_inode);
327 if (!error)
328 goto retry_deleg;
330 if (value != orig_value)
331 kfree(value);
333 return error;
335 EXPORT_SYMBOL_GPL(vfs_setxattr);
337 static ssize_t
338 xattr_getsecurity(struct mnt_idmap *idmap, struct inode *inode,
339 const char *name, void *value, size_t size)
341 void *buffer = NULL;
342 ssize_t len;
344 if (!value || !size) {
345 len = security_inode_getsecurity(idmap, inode, name,
346 &buffer, false);
347 goto out_noalloc;
350 len = security_inode_getsecurity(idmap, inode, name, &buffer,
351 true);
352 if (len < 0)
353 return len;
354 if (size < len) {
355 len = -ERANGE;
356 goto out;
358 memcpy(value, buffer, len);
359 out:
360 kfree(buffer);
361 out_noalloc:
362 return 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,
377 gfp_t flags)
379 const struct xattr_handler *handler;
380 struct inode *inode = dentry->d_inode;
381 char *value = *xattr_value;
382 int error;
384 error = xattr_permission(idmap, inode, name, MAY_READ);
385 if (error)
386 return error;
388 handler = xattr_resolve_name(inode, &name);
389 if (IS_ERR(handler))
390 return PTR_ERR(handler);
391 if (!handler->get)
392 return -EOPNOTSUPP;
393 error = handler->get(handler, dentry, inode, name, NULL, 0);
394 if (error < 0)
395 return error;
397 if (!value || (error > xattr_size)) {
398 value = krealloc(*xattr_value, error + 1, flags);
399 if (!value)
400 return -ENOMEM;
401 memset(value, 0, error + 1);
404 error = handler->get(handler, dentry, inode, name, value, error);
405 *xattr_value = value;
406 return error;
409 ssize_t
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))
416 return -EOPNOTSUPP;
418 handler = xattr_resolve_name(inode, &name);
419 if (IS_ERR(handler))
420 return PTR_ERR(handler);
421 if (!handler->get)
422 return -EOPNOTSUPP;
423 return handler->get(handler, dentry, inode, name, value, size);
425 EXPORT_SYMBOL(__vfs_getxattr);
427 ssize_t
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;
432 int error;
434 error = xattr_permission(idmap, inode, name, MAY_READ);
435 if (error)
436 return error;
438 error = security_inode_getxattr(dentry, name);
439 if (error)
440 return error;
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,
446 size);
448 * Only overwrite the return value if a security module
449 * is actually active.
451 if (ret == -EOPNOTSUPP)
452 goto nolsm;
453 return ret;
455 nolsm:
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
467 * inode of @dentry.
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
473 * xattrs.
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.
482 ssize_t
483 vfs_listxattr(struct dentry *dentry, char *list, size_t size)
485 struct inode *inode = d_inode(dentry);
486 ssize_t error;
488 error = security_inode_listxattr(dentry);
489 if (error)
490 return error;
492 if (inode->i_op->listxattr) {
493 error = inode->i_op->listxattr(dentry, list, size);
494 } else {
495 error = security_inode_listsecurity(inode, list, size);
496 if (size && error > size)
497 error = -ERANGE;
499 return error;
501 EXPORT_SYMBOL_GPL(vfs_listxattr);
504 __vfs_removexattr(struct mnt_idmap *idmap, struct dentry *dentry,
505 const char *name)
507 struct inode *inode = d_inode(dentry);
508 const struct xattr_handler *handler;
510 if (is_posix_acl_xattr(name))
511 return -EOPNOTSUPP;
513 handler = xattr_resolve_name(inode, &name);
514 if (IS_ERR(handler))
515 return PTR_ERR(handler);
516 if (!handler->set)
517 return -EOPNOTSUPP;
518 return handler->set(handler, idmap, dentry, inode, name, NULL, 0,
519 XATTR_REPLACE);
521 EXPORT_SYMBOL(__vfs_removexattr);
524 * __vfs_removexattr_locked - set an extended attribute while holding the inode
525 * lock
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;
539 int error;
541 error = xattr_permission(idmap, inode, name, MAY_WRITE);
542 if (error)
543 return error;
545 error = security_inode_removexattr(idmap, dentry, name);
546 if (error)
547 goto out;
549 error = try_break_deleg(inode, delegated_inode);
550 if (error)
551 goto out;
553 error = __vfs_removexattr(idmap, dentry, name);
554 if (error)
555 return error;
557 fsnotify_xattr(dentry);
558 security_inode_post_removexattr(dentry, name);
560 out:
561 return error;
563 EXPORT_SYMBOL_GPL(__vfs_removexattr_locked);
566 vfs_removexattr(struct mnt_idmap *idmap, struct dentry *dentry,
567 const char *name)
569 struct inode *inode = dentry->d_inode;
570 struct inode *delegated_inode = NULL;
571 int error;
573 retry_deleg:
574 inode_lock(inode);
575 error = __vfs_removexattr_locked(idmap, dentry,
576 name, &delegated_inode);
577 inode_unlock(inode);
579 if (delegated_inode) {
580 error = break_deleg_wait(&delegated_inode);
581 if (!error)
582 goto retry_deleg;
585 return error;
587 EXPORT_SYMBOL_GPL(vfs_removexattr);
589 int import_xattr_name(struct xattr_name *kname, const char __user *name)
591 int error = strncpy_from_user(kname->name, name,
592 sizeof(kname->name));
593 if (error == 0 || error == sizeof(kname->name))
594 return -ERANGE;
595 if (error < 0)
596 return error;
597 return 0;
601 * Extended attribute SET operations
604 int setxattr_copy(const char __user *name, struct kernel_xattr_ctx *ctx)
606 int error;
608 if (ctx->flags & ~(XATTR_CREATE|XATTR_REPLACE))
609 return -EINVAL;
611 error = import_xattr_name(ctx->kname, name);
612 if (error)
613 return error;
615 if (ctx->size) {
616 if (ctx->size > XATTR_SIZE_MAX)
617 return -E2BIG;
619 ctx->kvalue = vmemdup_user(ctx->cvalue, ctx->size);
620 if (IS_ERR(ctx->kvalue)) {
621 error = PTR_ERR(ctx->kvalue);
622 ctx->kvalue = NULL;
626 return error;
629 static int do_setxattr(struct mnt_idmap *idmap, struct dentry *dentry,
630 struct kernel_xattr_ctx *ctx)
632 if (is_posix_acl_xattr(ctx->kname->name))
633 return do_set_acl(idmap, dentry, ctx->kname->name,
634 ctx->kvalue, ctx->size);
636 return vfs_setxattr(idmap, dentry, ctx->kname->name,
637 ctx->kvalue, ctx->size, ctx->flags);
640 int file_setxattr(struct file *f, struct kernel_xattr_ctx *ctx)
642 int error = mnt_want_write_file(f);
644 if (!error) {
645 audit_file(f);
646 error = do_setxattr(file_mnt_idmap(f), f->f_path.dentry, ctx);
647 mnt_drop_write_file(f);
649 return error;
652 /* unconditionally consumes filename */
653 int filename_setxattr(int dfd, struct filename *filename,
654 unsigned int lookup_flags, struct kernel_xattr_ctx *ctx)
656 struct path path;
657 int error;
659 retry:
660 error = filename_lookup(dfd, filename, lookup_flags, &path, NULL);
661 if (error)
662 goto out;
663 error = mnt_want_write(path.mnt);
664 if (!error) {
665 error = do_setxattr(mnt_idmap(path.mnt), path.dentry, ctx);
666 mnt_drop_write(path.mnt);
668 path_put(&path);
669 if (retry_estale(error, lookup_flags)) {
670 lookup_flags |= LOOKUP_REVAL;
671 goto retry;
674 out:
675 putname(filename);
676 return error;
679 static int path_setxattrat(int dfd, const char __user *pathname,
680 unsigned int at_flags, const char __user *name,
681 const void __user *value, size_t size, int flags)
683 struct xattr_name kname;
684 struct kernel_xattr_ctx ctx = {
685 .cvalue = value,
686 .kvalue = NULL,
687 .size = size,
688 .kname = &kname,
689 .flags = flags,
691 struct filename *filename;
692 unsigned int lookup_flags = 0;
693 int error;
695 if ((at_flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
696 return -EINVAL;
698 if (!(at_flags & AT_SYMLINK_NOFOLLOW))
699 lookup_flags = LOOKUP_FOLLOW;
701 error = setxattr_copy(name, &ctx);
702 if (error)
703 return error;
705 filename = getname_maybe_null(pathname, at_flags);
706 if (!filename) {
707 CLASS(fd, f)(dfd);
708 if (fd_empty(f))
709 error = -EBADF;
710 else
711 error = file_setxattr(fd_file(f), &ctx);
712 } else {
713 error = filename_setxattr(dfd, filename, lookup_flags, &ctx);
715 kvfree(ctx.kvalue);
716 return error;
719 SYSCALL_DEFINE6(setxattrat, int, dfd, const char __user *, pathname, unsigned int, at_flags,
720 const char __user *, name, const struct xattr_args __user *, uargs,
721 size_t, usize)
723 struct xattr_args args = {};
724 int error;
726 BUILD_BUG_ON(sizeof(struct xattr_args) < XATTR_ARGS_SIZE_VER0);
727 BUILD_BUG_ON(sizeof(struct xattr_args) != XATTR_ARGS_SIZE_LATEST);
729 if (unlikely(usize < XATTR_ARGS_SIZE_VER0))
730 return -EINVAL;
731 if (usize > PAGE_SIZE)
732 return -E2BIG;
734 error = copy_struct_from_user(&args, sizeof(args), uargs, usize);
735 if (error)
736 return error;
738 return path_setxattrat(dfd, pathname, at_flags, name,
739 u64_to_user_ptr(args.value), args.size,
740 args.flags);
743 SYSCALL_DEFINE5(setxattr, const char __user *, pathname,
744 const char __user *, name, const void __user *, value,
745 size_t, size, int, flags)
747 return path_setxattrat(AT_FDCWD, pathname, 0, name, value, size, flags);
750 SYSCALL_DEFINE5(lsetxattr, const char __user *, pathname,
751 const char __user *, name, const void __user *, value,
752 size_t, size, int, flags)
754 return path_setxattrat(AT_FDCWD, pathname, AT_SYMLINK_NOFOLLOW, name,
755 value, size, flags);
758 SYSCALL_DEFINE5(fsetxattr, int, fd, const char __user *, name,
759 const void __user *,value, size_t, size, int, flags)
761 return path_setxattrat(fd, NULL, AT_EMPTY_PATH, name,
762 value, size, flags);
766 * Extended attribute GET operations
768 static ssize_t
769 do_getxattr(struct mnt_idmap *idmap, struct dentry *d,
770 struct kernel_xattr_ctx *ctx)
772 ssize_t error;
773 char *kname = ctx->kname->name;
774 void *kvalue = NULL;
776 if (ctx->size) {
777 if (ctx->size > XATTR_SIZE_MAX)
778 ctx->size = XATTR_SIZE_MAX;
779 kvalue = kvzalloc(ctx->size, GFP_KERNEL);
780 if (!kvalue)
781 return -ENOMEM;
784 if (is_posix_acl_xattr(kname))
785 error = do_get_acl(idmap, d, kname, kvalue, ctx->size);
786 else
787 error = vfs_getxattr(idmap, d, kname, kvalue, ctx->size);
788 if (error > 0) {
789 if (ctx->size && copy_to_user(ctx->value, kvalue, error))
790 error = -EFAULT;
791 } else if (error == -ERANGE && ctx->size >= XATTR_SIZE_MAX) {
792 /* The file system tried to returned a value bigger
793 than XATTR_SIZE_MAX bytes. Not possible. */
794 error = -E2BIG;
797 kvfree(kvalue);
798 return error;
801 ssize_t file_getxattr(struct file *f, struct kernel_xattr_ctx *ctx)
803 audit_file(f);
804 return do_getxattr(file_mnt_idmap(f), f->f_path.dentry, ctx);
807 /* unconditionally consumes filename */
808 ssize_t filename_getxattr(int dfd, struct filename *filename,
809 unsigned int lookup_flags, struct kernel_xattr_ctx *ctx)
811 struct path path;
812 ssize_t error;
813 retry:
814 error = filename_lookup(dfd, filename, lookup_flags, &path, NULL);
815 if (error)
816 goto out;
817 error = do_getxattr(mnt_idmap(path.mnt), path.dentry, ctx);
818 path_put(&path);
819 if (retry_estale(error, lookup_flags)) {
820 lookup_flags |= LOOKUP_REVAL;
821 goto retry;
823 out:
824 putname(filename);
825 return error;
828 static ssize_t path_getxattrat(int dfd, const char __user *pathname,
829 unsigned int at_flags, const char __user *name,
830 void __user *value, size_t size)
832 struct xattr_name kname;
833 struct kernel_xattr_ctx ctx = {
834 .value = value,
835 .size = size,
836 .kname = &kname,
837 .flags = 0,
839 struct filename *filename;
840 ssize_t error;
842 if ((at_flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
843 return -EINVAL;
845 error = import_xattr_name(&kname, name);
846 if (error)
847 return error;
849 filename = getname_maybe_null(pathname, at_flags);
850 if (!filename) {
851 CLASS(fd, f)(dfd);
852 if (fd_empty(f))
853 return -EBADF;
854 return file_getxattr(fd_file(f), &ctx);
855 } else {
856 int lookup_flags = 0;
857 if (!(at_flags & AT_SYMLINK_NOFOLLOW))
858 lookup_flags = LOOKUP_FOLLOW;
859 return filename_getxattr(dfd, filename, lookup_flags, &ctx);
863 SYSCALL_DEFINE6(getxattrat, int, dfd, const char __user *, pathname, unsigned int, at_flags,
864 const char __user *, name, struct xattr_args __user *, uargs, size_t, usize)
866 struct xattr_args args = {};
867 int error;
869 BUILD_BUG_ON(sizeof(struct xattr_args) < XATTR_ARGS_SIZE_VER0);
870 BUILD_BUG_ON(sizeof(struct xattr_args) != XATTR_ARGS_SIZE_LATEST);
872 if (unlikely(usize < XATTR_ARGS_SIZE_VER0))
873 return -EINVAL;
874 if (usize > PAGE_SIZE)
875 return -E2BIG;
877 error = copy_struct_from_user(&args, sizeof(args), uargs, usize);
878 if (error)
879 return error;
881 if (args.flags != 0)
882 return -EINVAL;
884 return path_getxattrat(dfd, pathname, at_flags, name,
885 u64_to_user_ptr(args.value), args.size);
888 SYSCALL_DEFINE4(getxattr, const char __user *, pathname,
889 const char __user *, name, void __user *, value, size_t, size)
891 return path_getxattrat(AT_FDCWD, pathname, 0, name, value, size);
894 SYSCALL_DEFINE4(lgetxattr, const char __user *, pathname,
895 const char __user *, name, void __user *, value, size_t, size)
897 return path_getxattrat(AT_FDCWD, pathname, AT_SYMLINK_NOFOLLOW, name,
898 value, size);
901 SYSCALL_DEFINE4(fgetxattr, int, fd, const char __user *, name,
902 void __user *, value, size_t, size)
904 return path_getxattrat(fd, NULL, AT_EMPTY_PATH, name, value, size);
908 * Extended attribute LIST operations
910 static ssize_t
911 listxattr(struct dentry *d, char __user *list, size_t size)
913 ssize_t error;
914 char *klist = NULL;
916 if (size) {
917 if (size > XATTR_LIST_MAX)
918 size = XATTR_LIST_MAX;
919 klist = kvmalloc(size, GFP_KERNEL);
920 if (!klist)
921 return -ENOMEM;
924 error = vfs_listxattr(d, klist, size);
925 if (error > 0) {
926 if (size && copy_to_user(list, klist, error))
927 error = -EFAULT;
928 } else if (error == -ERANGE && size >= XATTR_LIST_MAX) {
929 /* The file system tried to returned a list bigger
930 than XATTR_LIST_MAX bytes. Not possible. */
931 error = -E2BIG;
934 kvfree(klist);
936 return error;
939 static
940 ssize_t file_listxattr(struct file *f, char __user *list, size_t size)
942 audit_file(f);
943 return listxattr(f->f_path.dentry, list, size);
946 /* unconditionally consumes filename */
947 static
948 ssize_t filename_listxattr(int dfd, struct filename *filename,
949 unsigned int lookup_flags,
950 char __user *list, size_t size)
952 struct path path;
953 ssize_t error;
954 retry:
955 error = filename_lookup(dfd, filename, lookup_flags, &path, NULL);
956 if (error)
957 goto out;
958 error = listxattr(path.dentry, list, size);
959 path_put(&path);
960 if (retry_estale(error, lookup_flags)) {
961 lookup_flags |= LOOKUP_REVAL;
962 goto retry;
964 out:
965 putname(filename);
966 return error;
969 static ssize_t path_listxattrat(int dfd, const char __user *pathname,
970 unsigned int at_flags, char __user *list,
971 size_t size)
973 struct filename *filename;
974 int lookup_flags;
976 if ((at_flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
977 return -EINVAL;
979 filename = getname_maybe_null(pathname, at_flags);
980 if (!filename) {
981 CLASS(fd, f)(dfd);
982 if (fd_empty(f))
983 return -EBADF;
984 return file_listxattr(fd_file(f), list, size);
987 lookup_flags = (at_flags & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
988 return filename_listxattr(dfd, filename, lookup_flags, list, size);
991 SYSCALL_DEFINE5(listxattrat, int, dfd, const char __user *, pathname,
992 unsigned int, at_flags,
993 char __user *, list, size_t, size)
995 return path_listxattrat(dfd, pathname, at_flags, list, size);
998 SYSCALL_DEFINE3(listxattr, const char __user *, pathname, char __user *, list,
999 size_t, size)
1001 return path_listxattrat(AT_FDCWD, pathname, 0, list, size);
1004 SYSCALL_DEFINE3(llistxattr, const char __user *, pathname, char __user *, list,
1005 size_t, size)
1007 return path_listxattrat(AT_FDCWD, pathname, AT_SYMLINK_NOFOLLOW, list, size);
1010 SYSCALL_DEFINE3(flistxattr, int, fd, char __user *, list, size_t, size)
1012 return path_listxattrat(fd, NULL, AT_EMPTY_PATH, list, size);
1016 * Extended attribute REMOVE operations
1018 static long
1019 removexattr(struct mnt_idmap *idmap, struct dentry *d, const char *name)
1021 if (is_posix_acl_xattr(name))
1022 return vfs_remove_acl(idmap, d, name);
1023 return vfs_removexattr(idmap, d, name);
1026 static int file_removexattr(struct file *f, struct xattr_name *kname)
1028 int error = mnt_want_write_file(f);
1030 if (!error) {
1031 audit_file(f);
1032 error = removexattr(file_mnt_idmap(f),
1033 f->f_path.dentry, kname->name);
1034 mnt_drop_write_file(f);
1036 return error;
1039 /* unconditionally consumes filename */
1040 static int filename_removexattr(int dfd, struct filename *filename,
1041 unsigned int lookup_flags, struct xattr_name *kname)
1043 struct path path;
1044 int error;
1046 retry:
1047 error = filename_lookup(dfd, filename, lookup_flags, &path, NULL);
1048 if (error)
1049 goto out;
1050 error = mnt_want_write(path.mnt);
1051 if (!error) {
1052 error = removexattr(mnt_idmap(path.mnt), path.dentry, kname->name);
1053 mnt_drop_write(path.mnt);
1055 path_put(&path);
1056 if (retry_estale(error, lookup_flags)) {
1057 lookup_flags |= LOOKUP_REVAL;
1058 goto retry;
1060 out:
1061 putname(filename);
1062 return error;
1065 static int path_removexattrat(int dfd, const char __user *pathname,
1066 unsigned int at_flags, const char __user *name)
1068 struct xattr_name kname;
1069 struct filename *filename;
1070 unsigned int lookup_flags;
1071 int error;
1073 if ((at_flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
1074 return -EINVAL;
1076 error = import_xattr_name(&kname, name);
1077 if (error)
1078 return error;
1080 filename = getname_maybe_null(pathname, at_flags);
1081 if (!filename) {
1082 CLASS(fd, f)(dfd);
1083 if (fd_empty(f))
1084 return -EBADF;
1085 return file_removexattr(fd_file(f), &kname);
1087 lookup_flags = (at_flags & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
1088 return filename_removexattr(dfd, filename, lookup_flags, &kname);
1091 SYSCALL_DEFINE4(removexattrat, int, dfd, const char __user *, pathname,
1092 unsigned int, at_flags, const char __user *, name)
1094 return path_removexattrat(dfd, pathname, at_flags, name);
1097 SYSCALL_DEFINE2(removexattr, const char __user *, pathname,
1098 const char __user *, name)
1100 return path_removexattrat(AT_FDCWD, pathname, 0, name);
1103 SYSCALL_DEFINE2(lremovexattr, const char __user *, pathname,
1104 const char __user *, name)
1106 return path_removexattrat(AT_FDCWD, pathname, AT_SYMLINK_NOFOLLOW, name);
1109 SYSCALL_DEFINE2(fremovexattr, int, fd, const char __user *, name)
1111 return path_removexattrat(fd, NULL, AT_EMPTY_PATH, name);
1114 int xattr_list_one(char **buffer, ssize_t *remaining_size, const char *name)
1116 size_t len;
1118 len = strlen(name) + 1;
1119 if (*buffer) {
1120 if (*remaining_size < len)
1121 return -ERANGE;
1122 memcpy(*buffer, name, len);
1123 *buffer += len;
1125 *remaining_size -= len;
1126 return 0;
1130 * generic_listxattr - run through a dentry's xattr list() operations
1131 * @dentry: dentry to list the xattrs
1132 * @buffer: result buffer
1133 * @buffer_size: size of @buffer
1135 * Combine the results of the list() operation from every xattr_handler in the
1136 * xattr_handler stack.
1138 * Note that this will not include the entries for POSIX ACLs.
1140 ssize_t
1141 generic_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size)
1143 const struct xattr_handler *handler, * const *handlers = dentry->d_sb->s_xattr;
1144 ssize_t remaining_size = buffer_size;
1146 for_each_xattr_handler(handlers, handler) {
1147 int err;
1149 if (!handler->name || (handler->list && !handler->list(dentry)))
1150 continue;
1151 err = xattr_list_one(&buffer, &remaining_size, handler->name);
1152 if (err)
1153 return err;
1156 return buffer_size - remaining_size;
1158 EXPORT_SYMBOL(generic_listxattr);
1161 * xattr_full_name - Compute full attribute name from suffix
1163 * @handler: handler of the xattr_handler operation
1164 * @name: name passed to the xattr_handler operation
1166 * The get and set xattr handler operations are called with the remainder of
1167 * the attribute name after skipping the handler's prefix: for example, "foo"
1168 * is passed to the get operation of a handler with prefix "user." to get
1169 * attribute "user.foo". The full name is still "there" in the name though.
1171 * Note: the list xattr handler operation when called from the vfs is passed a
1172 * NULL name; some file systems use this operation internally, with varying
1173 * semantics.
1175 const char *xattr_full_name(const struct xattr_handler *handler,
1176 const char *name)
1178 size_t prefix_len = strlen(xattr_prefix(handler));
1180 return name - prefix_len;
1182 EXPORT_SYMBOL(xattr_full_name);
1185 * simple_xattr_space - estimate the memory used by a simple xattr
1186 * @name: the full name of the xattr
1187 * @size: the size of its value
1189 * This takes no account of how much larger the two slab objects actually are:
1190 * that would depend on the slab implementation, when what is required is a
1191 * deterministic number, which grows with name length and size and quantity.
1193 * Return: The approximate number of bytes of memory used by such an xattr.
1195 size_t simple_xattr_space(const char *name, size_t size)
1198 * Use "40" instead of sizeof(struct simple_xattr), to return the
1199 * same result on 32-bit and 64-bit, and even if simple_xattr grows.
1201 return 40 + size + strlen(name);
1205 * simple_xattr_free - free an xattr object
1206 * @xattr: the xattr object
1208 * Free the xattr object. Can handle @xattr being NULL.
1210 void simple_xattr_free(struct simple_xattr *xattr)
1212 if (xattr)
1213 kfree(xattr->name);
1214 kvfree(xattr);
1218 * simple_xattr_alloc - allocate new xattr object
1219 * @value: value of the xattr object
1220 * @size: size of @value
1222 * Allocate a new xattr object and initialize respective members. The caller is
1223 * responsible for handling the name of the xattr.
1225 * Return: On success a new xattr object is returned. On failure NULL is
1226 * returned.
1228 struct simple_xattr *simple_xattr_alloc(const void *value, size_t size)
1230 struct simple_xattr *new_xattr;
1231 size_t len;
1233 /* wrap around? */
1234 len = sizeof(*new_xattr) + size;
1235 if (len < sizeof(*new_xattr))
1236 return NULL;
1238 new_xattr = kvmalloc(len, GFP_KERNEL_ACCOUNT);
1239 if (!new_xattr)
1240 return NULL;
1242 new_xattr->size = size;
1243 memcpy(new_xattr->value, value, size);
1244 return new_xattr;
1248 * rbtree_simple_xattr_cmp - compare xattr name with current rbtree xattr entry
1249 * @key: xattr name
1250 * @node: current node
1252 * Compare the xattr name with the xattr name attached to @node in the rbtree.
1254 * Return: Negative value if continuing left, positive if continuing right, 0
1255 * if the xattr attached to @node matches @key.
1257 static int rbtree_simple_xattr_cmp(const void *key, const struct rb_node *node)
1259 const char *xattr_name = key;
1260 const struct simple_xattr *xattr;
1262 xattr = rb_entry(node, struct simple_xattr, rb_node);
1263 return strcmp(xattr->name, xattr_name);
1267 * rbtree_simple_xattr_node_cmp - compare two xattr rbtree nodes
1268 * @new_node: new node
1269 * @node: current node
1271 * Compare the xattr attached to @new_node with the xattr attached to @node.
1273 * Return: Negative value if continuing left, positive if continuing right, 0
1274 * if the xattr attached to @new_node matches the xattr attached to @node.
1276 static int rbtree_simple_xattr_node_cmp(struct rb_node *new_node,
1277 const struct rb_node *node)
1279 struct simple_xattr *xattr;
1280 xattr = rb_entry(new_node, struct simple_xattr, rb_node);
1281 return rbtree_simple_xattr_cmp(xattr->name, node);
1285 * simple_xattr_get - get an xattr object
1286 * @xattrs: the header of the xattr object
1287 * @name: the name of the xattr to retrieve
1288 * @buffer: the buffer to store the value into
1289 * @size: the size of @buffer
1291 * Try to find and retrieve the xattr object associated with @name.
1292 * If @buffer is provided store the value of @xattr in @buffer
1293 * otherwise just return the length. The size of @buffer is limited
1294 * to XATTR_SIZE_MAX which currently is 65536.
1296 * Return: On success the length of the xattr value is returned. On error a
1297 * negative error code is returned.
1299 int simple_xattr_get(struct simple_xattrs *xattrs, const char *name,
1300 void *buffer, size_t size)
1302 struct simple_xattr *xattr = NULL;
1303 struct rb_node *rbp;
1304 int ret = -ENODATA;
1306 read_lock(&xattrs->lock);
1307 rbp = rb_find(name, &xattrs->rb_root, rbtree_simple_xattr_cmp);
1308 if (rbp) {
1309 xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1310 ret = xattr->size;
1311 if (buffer) {
1312 if (size < xattr->size)
1313 ret = -ERANGE;
1314 else
1315 memcpy(buffer, xattr->value, xattr->size);
1318 read_unlock(&xattrs->lock);
1319 return ret;
1323 * simple_xattr_set - set an xattr object
1324 * @xattrs: the header of the xattr object
1325 * @name: the name of the xattr to retrieve
1326 * @value: the value to store along the xattr
1327 * @size: the size of @value
1328 * @flags: the flags determining how to set the xattr
1330 * Set a new xattr object.
1331 * If @value is passed a new xattr object will be allocated. If XATTR_REPLACE
1332 * is specified in @flags a matching xattr object for @name must already exist.
1333 * If it does it will be replaced with the new xattr object. If it doesn't we
1334 * fail. If XATTR_CREATE is specified and a matching xattr does already exist
1335 * we fail. If it doesn't we create a new xattr. If @flags is zero we simply
1336 * insert the new xattr replacing any existing one.
1338 * If @value is empty and a matching xattr object is found we delete it if
1339 * XATTR_REPLACE is specified in @flags or @flags is zero.
1341 * If @value is empty and no matching xattr object for @name is found we do
1342 * nothing if XATTR_CREATE is specified in @flags or @flags is zero. For
1343 * XATTR_REPLACE we fail as mentioned above.
1345 * Return: On success, the removed or replaced xattr is returned, to be freed
1346 * by the caller; or NULL if none. On failure a negative error code is returned.
1348 struct simple_xattr *simple_xattr_set(struct simple_xattrs *xattrs,
1349 const char *name, const void *value,
1350 size_t size, int flags)
1352 struct simple_xattr *old_xattr = NULL, *new_xattr = NULL;
1353 struct rb_node *parent = NULL, **rbp;
1354 int err = 0, ret;
1356 /* value == NULL means remove */
1357 if (value) {
1358 new_xattr = simple_xattr_alloc(value, size);
1359 if (!new_xattr)
1360 return ERR_PTR(-ENOMEM);
1362 new_xattr->name = kstrdup(name, GFP_KERNEL_ACCOUNT);
1363 if (!new_xattr->name) {
1364 simple_xattr_free(new_xattr);
1365 return ERR_PTR(-ENOMEM);
1369 write_lock(&xattrs->lock);
1370 rbp = &xattrs->rb_root.rb_node;
1371 while (*rbp) {
1372 parent = *rbp;
1373 ret = rbtree_simple_xattr_cmp(name, *rbp);
1374 if (ret < 0)
1375 rbp = &(*rbp)->rb_left;
1376 else if (ret > 0)
1377 rbp = &(*rbp)->rb_right;
1378 else
1379 old_xattr = rb_entry(*rbp, struct simple_xattr, rb_node);
1380 if (old_xattr)
1381 break;
1384 if (old_xattr) {
1385 /* Fail if XATTR_CREATE is requested and the xattr exists. */
1386 if (flags & XATTR_CREATE) {
1387 err = -EEXIST;
1388 goto out_unlock;
1391 if (new_xattr)
1392 rb_replace_node(&old_xattr->rb_node,
1393 &new_xattr->rb_node, &xattrs->rb_root);
1394 else
1395 rb_erase(&old_xattr->rb_node, &xattrs->rb_root);
1396 } else {
1397 /* Fail if XATTR_REPLACE is requested but no xattr is found. */
1398 if (flags & XATTR_REPLACE) {
1399 err = -ENODATA;
1400 goto out_unlock;
1404 * If XATTR_CREATE or no flags are specified together with a
1405 * new value simply insert it.
1407 if (new_xattr) {
1408 rb_link_node(&new_xattr->rb_node, parent, rbp);
1409 rb_insert_color(&new_xattr->rb_node, &xattrs->rb_root);
1413 * If XATTR_CREATE or no flags are specified and neither an
1414 * old or new xattr exist then we don't need to do anything.
1418 out_unlock:
1419 write_unlock(&xattrs->lock);
1420 if (!err)
1421 return old_xattr;
1422 simple_xattr_free(new_xattr);
1423 return ERR_PTR(err);
1426 static bool xattr_is_trusted(const char *name)
1428 return !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN);
1432 * simple_xattr_list - list all xattr objects
1433 * @inode: inode from which to get the xattrs
1434 * @xattrs: the header of the xattr object
1435 * @buffer: the buffer to store all xattrs into
1436 * @size: the size of @buffer
1438 * List all xattrs associated with @inode. If @buffer is NULL we returned
1439 * the required size of the buffer. If @buffer is provided we store the
1440 * xattrs value into it provided it is big enough.
1442 * Note, the number of xattr names that can be listed with listxattr(2) is
1443 * limited to XATTR_LIST_MAX aka 65536 bytes. If a larger buffer is passed
1444 * then vfs_listxattr() caps it to XATTR_LIST_MAX and if more xattr names
1445 * are found it will return -E2BIG.
1447 * Return: On success the required size or the size of the copied xattrs is
1448 * returned. On error a negative error code is returned.
1450 ssize_t simple_xattr_list(struct inode *inode, struct simple_xattrs *xattrs,
1451 char *buffer, size_t size)
1453 bool trusted = ns_capable_noaudit(&init_user_ns, CAP_SYS_ADMIN);
1454 struct simple_xattr *xattr;
1455 struct rb_node *rbp;
1456 ssize_t remaining_size = size;
1457 int err = 0;
1459 err = posix_acl_listxattr(inode, &buffer, &remaining_size);
1460 if (err)
1461 return err;
1463 read_lock(&xattrs->lock);
1464 for (rbp = rb_first(&xattrs->rb_root); rbp; rbp = rb_next(rbp)) {
1465 xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1467 /* skip "trusted." attributes for unprivileged callers */
1468 if (!trusted && xattr_is_trusted(xattr->name))
1469 continue;
1471 err = xattr_list_one(&buffer, &remaining_size, xattr->name);
1472 if (err)
1473 break;
1475 read_unlock(&xattrs->lock);
1477 return err ? err : size - remaining_size;
1481 * rbtree_simple_xattr_less - compare two xattr rbtree nodes
1482 * @new_node: new node
1483 * @node: current node
1485 * Compare the xattr attached to @new_node with the xattr attached to @node.
1486 * Note that this function technically tolerates duplicate entries.
1488 * Return: True if insertion point in the rbtree is found.
1490 static bool rbtree_simple_xattr_less(struct rb_node *new_node,
1491 const struct rb_node *node)
1493 return rbtree_simple_xattr_node_cmp(new_node, node) < 0;
1497 * simple_xattr_add - add xattr objects
1498 * @xattrs: the header of the xattr object
1499 * @new_xattr: the xattr object to add
1501 * Add an xattr object to @xattrs. This assumes no replacement or removal
1502 * of matching xattrs is wanted. Should only be called during inode
1503 * initialization when a few distinct initial xattrs are supposed to be set.
1505 void simple_xattr_add(struct simple_xattrs *xattrs,
1506 struct simple_xattr *new_xattr)
1508 write_lock(&xattrs->lock);
1509 rb_add(&new_xattr->rb_node, &xattrs->rb_root, rbtree_simple_xattr_less);
1510 write_unlock(&xattrs->lock);
1514 * simple_xattrs_init - initialize new xattr header
1515 * @xattrs: header to initialize
1517 * Initialize relevant fields of a an xattr header.
1519 void simple_xattrs_init(struct simple_xattrs *xattrs)
1521 xattrs->rb_root = RB_ROOT;
1522 rwlock_init(&xattrs->lock);
1526 * simple_xattrs_free - free xattrs
1527 * @xattrs: xattr header whose xattrs to destroy
1528 * @freed_space: approximate number of bytes of memory freed from @xattrs
1530 * Destroy all xattrs in @xattr. When this is called no one can hold a
1531 * reference to any of the xattrs anymore.
1533 void simple_xattrs_free(struct simple_xattrs *xattrs, size_t *freed_space)
1535 struct rb_node *rbp;
1537 if (freed_space)
1538 *freed_space = 0;
1539 rbp = rb_first(&xattrs->rb_root);
1540 while (rbp) {
1541 struct simple_xattr *xattr;
1542 struct rb_node *rbp_next;
1544 rbp_next = rb_next(rbp);
1545 xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1546 rb_erase(&xattr->rb_node, &xattrs->rb_root);
1547 if (freed_space)
1548 *freed_space += simple_xattr_space(xattr->name,
1549 xattr->size);
1550 simple_xattr_free(xattr);
1551 rbp = rbp_next;